devrelshow episode 45 - Deepa Subramanian - CourtneyR.dev's 2nd Brain

devrelshow episode 45 - Deepa Subramanian 🔗 **Source:** [devrelshow episode 45 - Deepa Subramanian](https://www.youtube.com/watch?v=fXJW_zEeZco) > [!note]+ Document Notes > tags: [devrel, community-organizing, ai-agents, google-developer-expert] > author: Frederic Harper > source: DevRel Show > reference: https://www.youtube.com/watch?v=fXJW_zEeZco > date created: 2024-12-28 17:32 > > # 🏷️ Title: DevRel Show Episode 45 - Community Leadership and AI Code Review with Deepa Subramanian >  ## 🧰 Action Inbox - [H1] Apply to Google Developer Expert program - requires invitation and community contribution #gde - [H2] Start local meetup or join existing GDG chapter - build community connections #community - [H3] Implement multi-agent AI code review pipeline - use different models for cost optimization #ai ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ### 🔗 References - [DevRel Show](https://devrel.show) - [GDG San Jose](https://gdg.community.dev/gdg-san-jose/) - [Deepa's LinkedIn](https://linkedin.com/in/deepa-subramanian) ## 📝 Summary - Episode 45 of DevRel Show featuring Deepa Subramanian discussing community organizing and AI-powered code review - Deepa shares her journey from non-traditional tech background to Google Developer Expert and GDG organizer - Demonstrates multi-agent AI system for automated code review using OpenAI and Anthropic models - Emphasizes importance of human oversight in AI processes and community building through consistent effort - Key takeaway: Community growth requires patience, collaboration, and authentic contribution over time ## 📖 Main Content Deepa Subramanian represents the growing trend of non-traditional tech professionals becoming community leaders. Her transition from academia (PhD background) to tech demonstrates that diverse backgrounds bring valuable perspectives to developer relations and community organizing. The Google Developer Expert program requires invitation-based application with four interview stages: screening, community assessment, product interview, and Googler interview. Success depends on authentic community contribution rather than artificial metrics like AI-generated content. Her AI code review demo showcases practical applications of multi-model AI systems, using OpenAI GPT-4 for summarization and Anthropic Claude for detailed code review and fixing suggestions. The system identifies security vulnerabilities, performance issues, and accessibility problems while maintaining human oversight as a critical component. ## 🗒️ Notes #### ⏱️ [00:00](https://youtu.be/fXJW_zEeZco?t=0) — [H1] Episode Introduction and Technical Difficulties - Key idea: Show rescheduled from 3pm to 2pm Eastern, causing StreamYard complications - Evidence: "I scheduled it at 2pm but I did the promotion for 3pm" #### ⏱️ [02:30](https://youtu.be/fXJW_zEeZco?t=150) — [H2] Deepa's Background and Community Journey - Key idea: 15+ years industry experience with non-traditional path to tech - Evidence: "I come from a non-traditional route... I was a woman techmakers ambassador" #### ⏱️ [05:15](https://youtu.be/fXJW_zEeZco?t=315) — [H2] Starting GDG San Jose Community - Key idea: Started from scratch in 2021, now has 1,400+ members - Evidence: "Community is not built in a day. It takes consistent effort" #### ⏱️ [08:45](https://youtu.be/fXJW_zEeZco?t=525) — [H3] Teaching and Mentorship Motivation - Key idea: Personal experience lacking mentors drives desire to help others - Evidence: "I never had a mentor... if at all I could be somebody's mentor" #### ⏱️ [12:20](https://youtu.be/fXJW_zEeZco?t=740) — [H1] Google Developer Expert Program Details - Key idea: Four-stage application process requiring invitation and authentic community contribution - Evidence: "You need to pass the screening... four or five steps" #### ⏱️ [18:30](https://youtu.be/fXJW_zEeZco?t=1110) — [H2] Community Event Collaboration - Key idea: Co-hosting events with other chapters increases reach and reduces organizer burden - Evidence: "If anybody enables co-host option, it relays to every other chapter" #### ⏱️ [22:15](https://youtu.be/fXJW_zEeZco?t=1335) — [H3] Paper-Based Learning Preference - Key idea: Visual learning through handwritten notes improves comprehension - Evidence: "I still do write in paper and pencil... I draw stuff" #### ⏱️ [25:40](https://youtu.be/fXJW_zEeZco?t=1540) — [H2] Benefits of Expert Programs - Key idea: Programs provide early access to technology and direct product team feedback - Evidence: "You have access to information prior... you can suggest these things" #### ⏱️ [32:10](https://youtu.be/fXJW_zEeZco?t=1930) — [H1] AI Code Review Demo Introduction - Key idea: Multi-agent system mimics real-world CI/CD pipeline with AI integration - Evidence: "How do you add AI agent to the pipeline... three agents" #### ⏱️ [35:45](https://youtu.be/fXJW_zEeZco?t=2145) — [H2] Multi-Model AI Strategy - Key idea: Different AI models optimized for specific tasks (OpenAI for summarization, Claude for review) - Evidence: "OpenAI's model is cheaper and faster for summarization" #### ⏱️ [38:20](https://youtu.be/fXJW_zEeZco?t=2300) — [H2] Backend Security Issues Detection - Key idea: AI agents identify SQL injection vulnerabilities and exposed API keys - Evidence: "Hard-coded secret... SQL injection vulnerability" #### ⏱️ [42:30](https://youtu.be/fXJW_zEeZco?t=2550) — [H3] Frontend Accessibility and Internationalization - Key idea: AI reviews detect missing alt text and RTL language layout issues - Evidence: "Missing alt text... layout changes for RTL languages" #### ⏱️ [45:15](https://youtu.be/fXJW_zEeZco?t=2715) — [H1] Human Oversight Requirement - Key idea: AI hallucination and errors require human validation in all processes - Evidence: "Still requires human eyes... don't trust what the agent says" #### ⏱️ [48:30](https://youtu.be/fXJW_zEeZco?t=2910) — [H3] Newsletter and Future Plans - Key idea: Starting newsletter focused on AI platform architecture discussions - Evidence: "Just started a newsletter... focusing on AI platform and architecture" ## 📚 References - [GDG San Jose YouTube Channel](https://youtube.com/@gdgsanjose) - [Google Developer Expert Program](https://developers.google.com/community/experts) - [Women Techmakers](https://womentechmakers.com) - [Deepa's Medium Blog](https://medium.com/@deepa-subramanian) ## 🗣️ Paraphrase - Community building requires consistent effort over time, starting small and gradually growing through authentic engagement rather than expecting immediate large audiences - The Google Developer Expert program values genuine community contribution and technical expertise over artificial metrics, requiring invitation and multi-stage interviews - AI-powered code review systems should use multiple specialized models for different tasks while maintaining human oversight to prevent hallucination-related errors ## 📰 Newsletter 🚀 **Community Leadership Meets AI Innovation** This week's DevRel Show highlighted how non-traditional backgrounds can bring fresh perspectives to developer relations. Deepa Subramanian's journey from academia to Google Developer Expert demonstrates that authentic community contribution matters more than formal credentials. **Key Takeaway**: Multi-agent AI systems for code review show promise, but human oversight remains critical. Consider using different AI models for specialized tasks - OpenAI for summarization, Claude for detailed analysis - to optimize both cost and accuracy. **Practical Application**: Start small with community organizing. Online-first meetups reduce logistical complexity while building initial audience. Collaborate with existing communities through co-hosting to expand reach efficiently. **Resources**: [Learn WordPress](https://learn.wordpress.org/), [Developer Docs](https://developer.wordpress.org/), [WP-CLI](https://wp-cli.org/) ## ✅ Action Items #### DevRel - Developer advocacy: Create authentic content demonstrating real expertise rather than AI-generated materials - Education: Develop mentorship programs for non-traditional tech backgrounds entering developer relations - Developer experience: Implement multi-model AI systems for code review while maintaining human validation - Community management: Establish co-hosting partnerships with adjacent communities for expanded reach #### Community - Maintainers: Document multi-stage application processes for recognition programs - Contributors: Focus on consistent, authentic contribution over artificial engagement metrics - Ecosystem: Develop collaboration frameworks for cross-community event organization ## 🧪 Hypotheses - Multi-model AI systems provide better cost-performance ratios than single-model approaches for complex code review tasks - Community organizers from non-traditional backgrounds bring unique perspectives that enhance developer engagement - Co-hosted virtual events generate higher attendance and engagement than single-community events - Paper-based learning techniques improve technical concept retention compared to digital-only methods ### 📊 Methodology - A/B test single-model vs multi-model AI code review systems measuring accuracy and cost - Survey community members on organizer background diversity impact on event quality and inclusion - Compare attendance and engagement metrics between solo and co-hosted virtual events ### ❓ Questions - How can recognition programs like GDE scale while maintaining quality standards and authentic contribution requirements? - What combination of AI models provides optimal cost-effectiveness for different code review tasks? - How do non-traditional career paths contribute unique value to developer relations roles? ### 🔮 Future Research - Impact of human oversight percentage on AI code review accuracy and team productivity - Correlation between community organizer diversity and long-term member retention rates - Effectiveness of paper-based vs digital learning methods for technical skill development ### 💡 Implications - Practice: Implement multi-stage validation processes for AI-assisted code review - Policy: Develop guidelines requiring human oversight in AI-powered development workflows - Education: Create pathways for non-traditional backgrounds entering developer relations - Funding: Budget for multiple AI model subscriptions rather than single-provider solutions ### 🧠 Learning Styles Assessment Reading level estimate: 12th grade #### 👀 Visual learners - Present: Code examples, architecture diagrams in demo - Improve: Add flowcharts showing GDE application process, community growth metrics #### 🔊 Aural learners - Present: Conversational interview format, verbal explanations - Improve: Provide audio-only version with enhanced sound quality #### 🗣️ Verbal learners - Present: Detailed explanations, analogies, technical terminology definitions - Improve: Add glossary of community organizing terms, more real-world examples #### 🧩 Kinesthetic learners - Present: Live code demo, step-by-step technical walkthrough - Improve: Provide downloadable code samples, interactive setup guides #### 🧠 Logical learners - Present: Four-stage GDE process, systematic approach to AI model selection - Improve: Add decision trees for model selection, ROI calculations for community events #### 🧑‍🤝‍🧑 Social learners - Present: Community collaboration examples, co-hosting strategies - Improve: Include discussion prompts, community building checklists #### 🧘‍♀️ Solitary learners - Present: Individual learning journey narrative, self-directed skill development - Improve: Add reflection questions, personal development frameworks #### 🌿 Naturalistic learners - Present: Organic community growth analogies, ecosystem thinking - Improve: Connect to broader tech ecosystem patterns, sustainable growth models ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Conversational format with clear speaker identification - Improve: Add detailed alt text for code demonstrations, structured headings #### Keyboard navigation - Present: Standard video player controls - Improve: Provide transcript with timestamp navigation links #### Contrast and readability - Present: Clear video quality, readable code samples - Improve: Increase code font size, ensure high contrast in IDE themes #### Media and captions - Present: Full transcript available, clear audio - Improve: Add professional captions with speaker identification, code snippet descriptions ### 🌍 DEIB Assessment #### Cultural breadth - Present: Global community perspective, international collaboration examples - Improve: Include more diverse geographic examples, multilingual community strategies #### Inclusive language - Present: Gender-neutral terms, welcoming community language - Improve: Define technical jargon consistently, avoid cultural idioms #### Neurodiversity support - Present: Clear structure, consistent pacing - Improve: Provide content warnings for rapid technical demonstrations, offer alternative formats #### Representation - Present: Non-traditional career path representation, diverse professional backgrounds - Improve: Include more varied identity representation in examples, accessibility considerations ### ✅ Logical Validity #### Strong reasoning - Deductive strength: GDE program requirements → community contribution necessity → authentic engagement focus - Inductive strength: Individual community success → broader community organizing principles - Causal reasoning: Consistent effort → community growth → sustained engagement - Analogical strength: Academic mentorship needs → tech community mentorship needs #### Fallacies - Survivorship bias: Success story focus without failure case analysis - could include balanced perspectives on challenges - Appeal to authority: GDE/Google credibility emphasis - balanced with independent validation of strategies #### Tone of voice - Overall tone: Enthusiastic and encouraging with authentic vulnerability about challenges and mistakes ### 🛡️ DARVO Analysis - Deny: No denial patterns present - Attack: No attacking behavior observed - Reverse victim/offender: Not applicable to this content - Emotional cues and rhetoric: Positive, supportive language encouraging authentic participation and learning ## 🎯 Conclusion Community building and AI implementation both require patience, authenticity, and human oversight. Deepa's journey demonstrates that non-traditional backgrounds bring valuable perspectives to developer relations, while her AI demo shows the promise and limitations of current automated systems. Success in both domains comes from consistent effort, genuine contribution, and maintaining human judgment in automated processes. ## 🧱 Refactor Appendix ### 📚 Literature Split Pad Purpose: Stage CEE units that become Literature notes. Split level: H3 ### [H1] Google Developer Expert programs require authentic community contribution > [!quote] DevRel Show Episode 45 by [[Frederic Harper]] via [[YouTube]]. Published 2024. Accessed 2024-12-28. > "Please do not do AI generated videos on YouTube and say that I've created a channel. They want to see your face. They want to see how authentic you are." Authenticity over artificial metrics in recognition programs. [view highlight](https://youtu.be/fXJW_zEeZco#H1) #xref #articleseed ID:rw_gde_authentic ### [H2] Multi-model AI systems optimize cost and performance > [!quote] DevRel Show Episode 45 > "OpenAI's model is cheaper and faster for summarization. So that's why I chose this." Strategic model selection based on task-specific strengths. [view highlight](https://youtu.be/fXJW_zEeZco#H2) #xref ID:rw_ai_multimodel ### [H3] Community growth requires consistent long-term effort > [!quote] DevRel Show Episode 45 > "Community is not built in a day. It --- ## Full Transcript [00:31] Hello, hello lovely people. Welcome to [00:34] another episode of the Devro Show. I [00:36] think it's episode number 45. And uh [00:40] hopefully uh you're able to join us [00:42] right now. Uh that was not my best [00:45] management of an episode of the show. [00:47] You know, usually the show goes at 3 [00:49] p.m. Eastern time and I scheduled it at [00:52] 2 p.m. But I I did the promotion for 3 [00:54] p.m. And uh like because of that, I had [00:57] to reschedu things in StreamYard and I [01:00] had to delete some stuff on LinkedIn and [01:02] YouTube and recreate it. So hopefully [01:03] you find the right link to uh join us [01:06] live today. In uh worst case scenario, [01:09] uh the lovely thing with those webinars [01:10] is that there's always the recording [01:12] available. So, you're going to be able [01:13] to uh watch the show uh after if you're [01:16] not able to join us live. But hopefully, [01:18] it's always fun to have people live uh [01:20] with questions and comments. Uh but you [01:22] know, as I said, sometimes even when [01:24] there are people live, I didn't tell [01:25] you, but like sometimes we have people [01:27] live and it's just like nobody's there [01:29] because nobody's asking question, [01:31] nobody's commenting. And sometime we get [01:33] 10,000 of those. So it really depends on [01:36] like you know how's the sun today, [01:38] what's the weather, what's the alignment [01:40] of the stars. Like there's no real way [01:42] to uh understand exactly when we have a [01:44] lot more activities or not. But uh [01:46] always happy to do the Devrell show [01:48] especially since uh I started to travel [01:50] again for work. They're a little less uh [01:53] every other week like they used to be. [01:55] So sometimes there's a little more time [01:56] between each other. But uh I continue to [01:59] do those. Still really excited. I have a [02:01] really long list of uh guests in the [02:04] wait list. So uh they won't stop uh [02:07] tomorrow. So again really happy to be [02:09] here. It's always fun because I always [02:10] have amazing guests. And today with me I [02:13] have uh Deepa. How are you? [02:15] >> Hey, thank you so much Brad. I'm doing [02:18] very well and no worries about the whole [02:21] no Streamyard YouTube matchup thing. [02:23] You're totally fine. [02:25] >> Wonderful. Wonderful. So thanks for [02:26] joining. Uh we waited a long time. Uh [02:30] sorry. [02:30] >> So sorry. Uh do I have to join the chat [02:33] here? Uh so [02:35] >> no, you're good. [02:36] >> I'm good. Okay. Thank you. [02:37] >> No, no, you're good. You're good. If [02:38] there is anybody uh you know commenting, [02:40] I will share them on the screen. Uh so [02:43] yeah, no, you're good here. But uh no, [02:45] thanks for uh thanks for being a guest [02:47] on the show. Uh now that's probably the [02:49] story. Unfortunately, everybody who [02:51] wants to be uh on the show wants to be a [02:53] guest uh because there's a long wait [02:55] list and I don't do them as often as I [02:57] would like to because of my schedule. Uh [02:59] it takes a little bit of time. So, I [03:01] think we talked about this a couple of [03:02] months ago if a year ago, but finally I [03:05] was able to have you on the show. So, [03:07] first question for you. Uh who are you? [03:11] >> That's a good question. So, well, um for [03:14] people I'm I'm not new to the community [03:16] actually. Um, for people who know me, [03:19] uh, I'm Deepa. I have been in the [03:20] industry for 15 plus years, but I also [03:23] run a lot of communities around my local [03:25] area. I live in San Jose, California. I [03:28] run GDG San Jose. And I there was a part [03:32] that, um, before I even started out with [03:35] San Jose, um, I was a woman techmakers [03:38] ambassador. Now it's been moved on to [03:40] Technovision. Um so yeah uh been in [03:44] around with the community for more than [03:45] five years now. So professionally I've [03:49] been working for more than 15 plus [03:51] years. Uh I have come a long long way. [03:54] So yeah my previous my last job was at [03:57] Adobe and then um I my career path was [04:01] totally a different one a [04:03] non-traditional one as well for people [04:06] who do not know me. I come. Yeah, it's [04:09] it's a totally different route that I [04:11] took uh back in the days. So, but I'm [04:14] excited and I'm also happy that I made [04:17] that decision. So, yeah, that's me. [04:20] >> No, that's great. [04:20] >> If you want to follow and if you want to [04:22] follow me everything, Fred will post it [04:25] at the end. [04:25] >> I will share a lot of link actually. Uh [04:28] you just remembered me that I have a lot [04:29] of links to share and I for whatever [04:31] reason I was so into what you were [04:32] saying that I forgot about those. But [04:34] yeah, during the show I'm going to share [04:35] a lot of links uh where uh places where [04:37] you can connect with depot and uh really [04:40] happy uh actually I have a little bit [04:42] memory so I don't know for how long we [04:44] know each other. I know I've I've spoke [04:46] once at your user group and it's also [04:48] part of the reason why um I was happy to [04:51] have you outside of the fact that you're [04:52] an amazing person [04:55] but you know like it's a dev show. So, [04:57] usually the guests I have on the show, [04:59] uh, they usually have the title [05:00] developer advocate or developer [05:03] relations something or technical [05:04] evangelist or like the kind of title [05:06] where like you your full time is doing [05:08] that. But, uh, once in a while because I [05:11] don't want to restrict myself. Uh, I [05:13] want also people that may not have like [05:15] the official title but still do some or [05:19] many of the things that developer [05:21] advocates do, developer relation. And [05:23] part of it is uh your involvement in the [05:26] community is the fact that you're a [05:28] community organizer as you said like the [05:30] GDG uh San Jose and uh for me those are [05:34] some of the things that usually [05:36] depending on a company depending on your [05:38] role uh you know uh developer advocates [05:40] may do uh in in in their day-to-day job. [05:43] So I invited you today as like a [05:46] community community organizer. So I [05:48] would like to you know uh start with [05:50] that uh you know [05:52] >> why did you start a community? Why are [05:54] you doing that? [05:55] >> What are you doing? [05:56] >> Um so the story goes behind in 2021. Um [06:01] so um the thing was I started off with a [06:04] whole community thing. Google gave me an [06:05] opportunity to talk at um uh you know [06:10] talk uh be in a program called as women [06:12] developer academy back in 2021 and they [06:14] were training about 30 women for this [06:17] academy and I was chosen it was there [06:20] were 300 applicants and I was chosen one [06:22] so uh amongst the 30 people so that's [06:25] how I got a step or I came to even know [06:28] about this community um community as a [06:31] whole what I could contribute to the [06:33] community I [06:35] you know personally I'm very close to my [06:37] I know teaching is something that is [06:38] very close to my heart and tech is [06:41] something that I just you know I chose a [06:44] way to transition to and when I was uh [06:48] you know getting into tech I have had my [06:51] own struggles to grow and I I never had [06:56] a mentor so that's I was if if at all I [07:00] could do and do something and if at all [07:02] I could be somebody's mentor. Um that's [07:06] that's how it all started. So I I do [07:08] love speaking. Um this was not in [07:10] technique uh tech part of my uh life but [07:13] then I've I've done a lot of um when I [07:15] was uh in my PhD program. I have given [07:18] talks, I've written research work, [07:21] papers and all those other stuff. So [07:23] public speaking was never a fear or [07:26] anything like that. But how do I bring [07:28] it to the you know bring bring it into [07:32] tech was a was something that I did not [07:34] know. When Google gave me this [07:35] opportunity in the WBA program I stepped [07:38] onto it right away. And then they said [07:41] hey do you want to do events? Do you [07:43] want to be speaking somewhere? I'm like [07:45] oh I could I want to do this. So they [07:48] said that you could teach you could [07:49] inspire people with what you know. So I [07:53] was not very sure how people would take [07:56] me at the beginning or would they even [07:58] accept me from coming from a [08:00] non-traditional [08:02] uh I was a bit hesitant. So then once I [08:04] started doing it I'm like okay this is [08:06] fun. So and when I started uh speaking [08:11] as well as this is when they said hey [08:13] can you also do you know events? So I [08:17] create I started this GDG San Jose in [08:20] California. um in 2021 I guess. So yeah [08:26] and I started off from scratch. So for [08:29] people out here your community is not [08:31] built in a day. It takes consistent [08:34] effort and today I'm very happy that [08:38] there are about 1,400 members in the [08:40] community that I run it. So it it does [08:43] but last year was a little bit of um um [08:46] little bit I was not very active but [08:50] this year I I will be active so that's [08:53] something we are you know I'm planning [08:55] to do a lot of events during this year [08:58] and Google as a whole they you know they [09:00] have campaigns for every quarter so um [09:05] earlier it used to be about four events [09:08] per year now the count has become uh [09:11] six. So that's something that you know I [09:15] have to schedule it accordingly and do [09:17] it. And this this is how I started and [09:21] when I say events I've done about I [09:25] don't have a count I think last year I [09:27] think 60 to 70 events uh together but [09:30] then I also co-host with other events [09:32] too. So if anybody is doing across in [09:35] the US or anywhere in the globe and they [09:38] said hey do you want to co co-host with [09:40] uh with my chapter. So I co-host with [09:42] them too. So I'm invol I know that there [09:44] are um couple of things one in Sudan and [09:48] uh there are a couple of other chapters [09:50] they ping me the organizers ping me and [09:52] ask hey can you co-host it I'm open to [09:54] but provided it's in the time zone and [09:57] the right time it's there. So um so you [10:00] collaborate too that way. And the third [10:03] part of the community was um I was while [10:07] I was also speaking and giving talks um [10:11] I became a Google developer expert too. [10:15] So this is something uh yeah I did not [10:18] say that to you Fred because [10:21] >> last year yeah last year my uh and [10:24] Google you know for any events that you [10:26] do any talks that you do as a GDE you we [10:30] have a reporting system so uh I have to [10:33] report it but then in in doing events is [10:36] pretty easy once you're done you have to [10:38] wrap up the event and say that hey this [10:40] and it automatically uh you know it'll [10:42] tell you how many if it's an online [10:43] event it'll that tell you how many [10:45] online attendees were there. Um, so the [10:49] event wraps up and it automatically does [10:52] it. However, uh, in the GD program, you [10:56] have to go and, uh, probably put in all [10:59] your details and that's where I forgot [11:02] last year. I would be very and I kind of [11:05] lost it. Um and I had to talk to the G [11:09] uh this and I it just got reinstated [11:13] saying that hey so I said I'm sorry I [11:15] forgot to do my activity. These are the [11:17] things that I did because last year was [11:19] a whole mess. Um I try to u do my [11:24] founder job. I try to bring in a youth [11:27] tech in there. We gave seven weeks of [11:30] free summer program for high school kids [11:33] and I was super duper busy with that. [11:36] So, I totally forgot, but that was my [11:38] bad. Um, so this year I'm starting to do [11:42] something more to the community. Let's [11:45] see how it goes. So, [11:47] >> yeah, it's it's it's fun. Actually, I [11:49] love doing this because it connects me [11:52] to um where I want, you know, how if if [11:57] at all, if you ask me 20 years back, if [11:59] somebody could have helped me, I would [12:00] have been in a better place today. I [12:02] could have had a proper mentor for eight [12:04] years of my life. I experimented with [12:06] different roles trying to understand the [12:09] entire systems you know so but then you [12:12] know I don't regret it everything was a [12:14] learning everything was a very very [12:16] calculated professional move that I did [12:19] uh to learn a lot of stuff uh and I'm [12:23] still learning I'm not I'm not an expert [12:25] in anything I'm still learning we're all [12:27] learning I guess and that's what keeps [12:29] uh keeps u you know keeps me awake at [12:32] night so if there's something new I just [12:34] jump jump onto it. I'm like, what is [12:36] this? How do I make it? And for me, I'm [12:38] a very, very visual learner at first. [12:40] So, I need to put my pieces. I need to [12:43] write up. I still do write in paper and [12:45] pencil. If I don't understand stuff, I [12:47] have to sit and think, oh, what is this? [12:49] How do you do this? I am very visual. I [12:51] draw stuff. I that's how I understand. [12:54] And once I understand it, it's very easy [12:56] for me to communicate with people, too. [12:59] >> Yeah. There's uh about that that pen and [13:01] paper. Uh I mean, I I said that one. I [13:04] do this once in a while too and I I just [13:06] felt like I stone hold when I was saying [13:08] that but like there's research about uh [13:10] the fact that and and I can't remember [13:12] exactly but it was about the fact that [13:14] like when you writing down like with the [13:16] proper pen uh on paper and I don't think [13:20] there there's the same effect like with [13:21] a iPad or something but I don't know I [13:23] can't remember if it's related to like [13:25] the feeling of paper the fact that your [13:27] brain first writing is something it [13:29] helps your brain remember better or [13:31] learn better instead of like typing on [13:33] keyboard and I I wish I would remember [13:37] right now why it was the case but uh I [13:40] mean go on Google search just trust me [13:42] uh people think there's research about [13:44] the fact that like when you write on [13:45] paper but like on real paper uh there's [13:48] something about it that makes your brain [13:50] do better at learning and and memorizing [13:54] and understanding stuff. So it's why [13:56] like many people outside even outside of [13:58] text like when you want as you said when [13:59] you want to learn a concept sometimes [14:01] it's easier to go on a even on the board [14:03] even if it's not a real paper but like [14:04] writing on the board or this is why like [14:07] psychologists tell you like hey if you [14:09] write a journal like like buy a paper [14:11] one because that helps you probably [14:13] better process your emotion and stuff [14:14] like that. So I think it's just a great [14:16] uh a great way to uh you know try to [14:19] understand things which is also a [14:21] benefit of working in tech. I feel like [14:23] one of the thing I like working in tech [14:26] uh is that you constantly need to learn [14:28] because it's going so fast that if you [14:30] don't continually learn new stuff, [14:33] you're just going to be left behind uh [14:35] because going so fast, especially right [14:37] now with AI and all that stuff. So I [14:39] like it. There's so many things I would [14:41] like to say because you said a lot of [14:43] things, but uh let me try to to [14:45] cherrypick some of those things. Uh the [14:47] first one um would love to talk uh a [14:50] little bit about like uh actually you [14:53] said it so maybe you don't have that [14:54] much more to say but uh Google uh the [14:56] Google developer expert uh recognition [14:59] from Google because um I know a couple [15:01] people that are uh expert on Google but [15:04] I don't think it was ever a topic of [15:06] discussion in Deal show and I love those [15:08] kind of like expert MVP or recognition [15:12] program like you you're an expert in [15:14] specific dominant or you're recognized [15:16] by the company by being what I would [15:18] call a virtual evangelist for the [15:20] company like so not [15:21] >> working as an employee but still improve [15:24] the community improve the the whole [15:26] ecosystem so um h how is it work I know [15:30] you already said a couple things [15:32] >> so it it it does it does require [15:34] somebody to invite you okay there is a [15:37] application process and even prior to [15:41] you applying you should have been [15:44] involved with the community Um if I [15:47] think I have a blog about it on my [15:49] Medium post um I think three years back [15:51] that I wrote. If anybody is interested [15:54] just go read it or shoot me a message in [15:56] LinkedIn. I am very active there. So I [15:58] will I will tell you hey exactly what [16:00] you need to do. But you need to do uh [16:02] they they need to know how much have you [16:04] contributed to the community. Right? So [16:08] when I started up even before I applied [16:10] I think I was invited for a program [16:12] called as Google developer expert. um uh [16:16] road map it was for it was through women [16:19] techmakers um in 2022. [16:23] So um I was invited for this program [16:26] 2022 yes 2022. So I was invited for this [16:29] program um prior to that I have given [16:32] you know few talks already. I was around [16:35] the community. I was also doing events. [16:36] Events help you a lot because you [16:38] connect with a lot of people too. Yes. I [16:40] I never had the intention to become an [16:43] expert. However, uh they gave me a ch [16:47] chance. I'm like, "Oh, you have you know [16:49] because I love teaching and I I love [16:51] speaking. Um I gave talks around a lot [16:55] of communities especially in the Google [16:57] those back in the days it was called [16:58] this Google developer um campus groups. [17:02] I now it is called as GDSE that's that [17:05] was uh the term that they use. Now it's [17:07] all Google developer groups on campus. [17:10] So I have done a lot of onampus stuff [17:13] either virtual or going there and [17:14] speaking as well. So um my because I did [17:21] all the work prior uh if you go to my [17:24] website and there is a talks page I [17:27] think and it gives all the information [17:29] of what I started. I did a lot of [17:31] mistake. I was not super duper experty. [17:36] I was scared to at the beginning but now [17:39] it's like okay even if I make mistakes [17:41] you know I just say oh yeah I made a [17:43] mistake so I move on [17:44] >> that's fine [17:44] >> so um initially I was very hesitant [17:47] everything had to work perfectly if I [17:49] even fumble I I used to you know do a [17:52] lot of judging self judging myself um [17:56] however it's just that somebody thought [17:58] oh you're doing very well why don't you [18:00] apply for this I'm like okay so the GDE [18:03] road you know the road to GDE program I [18:06] was also involved and this is uh I think [18:09] one person among globally was chosen. M [18:12] >> so yeah uh I was so happy that you know [18:15] it's like somebody recognized oh I was [18:17] thinking oh maybe you know somebody is [18:20] recognizing I I think I do have some [18:22] skills right so coming from a [18:24] non-traditional area so I felt that but [18:28] then you know what when somebody calls [18:29] you and says that hey you are you know [18:32] you are you know you are visible you are [18:36] doing something to the community and we [18:37] like you too right [18:39] >> so [18:41] uh I was in the program. So they do have [18:43] an application. Uh if you go to my blog, [18:45] I I put pictures of all the steps too. [18:47] There are four or five. I don't know it [18:49] might have changed. First thing you need [18:50] to pass the screening. So the p to pass [18:54] the screening itself um once it you pass [18:56] it. Um and there are two more three more [19:00] uh interviews. one with the community [19:02] person, one with the product person from [19:06] from the community itself and then the [19:08] last one would be with a Googler. So [19:11] there are four steps to it and the road [19:14] to GDE program helped me as how do I do [19:18] how do I make my application stronger? [19:21] What how do I do this? Right? I was all [19:23] over the place. So I came from web from [19:26] the web space. I love React. I love [19:29] Angular. you know I I kind of you know [19:31] because the time that I was working uh [19:34] there we had both Angular and React so I [19:37] was like oh no I find this you know and [19:40] any new thing excited me I'm like okay [19:42] let me try this and then you know that's [19:45] how they said no no no you cannot go [19:47] around this place just pick one [19:49] particular technology that you like I [19:51] said I don't want to be in one [19:53] particular framework I don't want to be [19:55] in super duper angular or I and I I was [19:59] Flutter is something totally you know [20:02] it's not going to be used in my in my [20:04] workplace so it'll be a totally [20:06] different domain for me or a different [20:08] platform for me. So I said I want to [20:10] remind in web but I don't want to pick [20:13] any frameworks per se. That's when um [20:16] they said okay why don't you go in web [20:18] technologies there are four options now [20:20] the categories are totally different now [20:22] everything is web AI um so uh I started [20:26] off with the web APIs the the uh [20:29] progressive web app that was the [20:32] category that was in there so I was [20:35] working on the browser APIs I did a lot [20:37] of talks around that I did a lot of [20:39] writing around that please don't you [20:41] know my previous blogs would have been [20:43] very very Um it it is all my [20:46] experimentation as because I first [20:50] >> and the best part is if you don't [20:51] understand something you try to learn it [20:54] and write a write a blog about it and [20:56] talk about it. That's that's the whole [20:59] point of me doing the same rinse and [21:01] repeat thing if any technology comes up. [21:05] So even now the the AI thing is a big [21:07] hype. I'm trying to do the same thing. [21:09] Okay, what did I do before? How could I [21:11] do differently this time? Right? So [21:14] that's one thing that um I did and then [21:18] the program um there there was a mock [21:21] application they showed it to us and I [21:22] had a mentor I I should uh I had couple [21:26] of mentors along the way one was uh [21:29] James um he was he is in um he's in UK [21:33] he James Milner that's his name he was [21:36] also he is still a web GDE and he was my [21:40] mentor and he was guiding me through [21:42] along the whole process [21:44] you should do this, you know, that [21:45] that's something that I have never had [21:47] in my career at all. [21:48] >> So, it's not that somebody handheld me. [21:51] He was just pointing me maybe, you know, [21:53] this could do better. This could do [21:54] better. So, I was like, oh, can I do [21:56] this? Can I do that, too? So, I was all [21:58] I I get very hyped up and if you if you [22:00] see me going, you know, going, [22:03] >> you [22:05] become a cartoon, so don't please don't [22:08] mind my facial expressions. However, I [22:10] get very excited about this. Um so I do [22:15] the I did this. So that's how the GDE [22:17] happened. Let's not talk about that. [22:19] That's that's totally a different event. [22:21] Coming to the events, events are [22:23] something that is that connected me to a [22:25] lot of people. So it does not happen [22:27] overnight. Like I said, um the GDG San [22:31] Jose LinkedIn page is created um by me. [22:35] So if you see that there are 500 people, [22:37] you can also follow me there. But the [22:39] best way is to go to GDG San Jose um you [22:43] know Google's page. So yeah that's the [22:45] page. So if you go there you can join us [22:48] and you will get all the updates [22:50] instantly even if the co-hosting ones [22:52] and you can pick and choose what events [22:54] works for you. Uh there there is a [22:56] campaign that's going to be coming this [22:58] I think starting March. Everybody I know [23:01] we are all working on the build with AI [23:03] thing. So it's going to be coming um [23:06] with new whatever new stuff being in the [23:09] community also gives you an access to [23:11] closer to technology whatever Google is [23:14] building um it gets you closer but [23:17] becoming a GDE gets even more closer [23:20] let's let's put it this way you you have [23:23] an access to um the information prior of [23:27] course you are on uh I'm on an NGA I [23:29] cannot talk about it uh however you can [23:32] suggest these things Hey, I tried this [23:34] thing. Maybe this feature could be, you [23:36] know, we want more of these feature or [23:38] we want more of this or maybe I'm having [23:40] issues and maybe you know all the [23:42] suggestions that you would give at any [23:44] you know you do that with any company [23:46] level too right it's similar things and [23:50] it's that's something [23:53] you know you you as a person grow along [23:56] with tech as well as helping others grow [23:59] as well. [24:01] >> Yeah. And know that's one of that's [24:02] that's one of benefit. It looks a little [24:04] bit like the other program I know being [24:06] part of or help to to manage or create [24:09] and um I see there's benefits from the [24:12] company but it always need to be a [24:13] win-win situation and and this is one of [24:15] the benefits where like you have access [24:17] to the product team you can give [24:18] feedback you have more information uh [24:20] hopefully there is a lot of so support [24:22] for you to doing those kind of things [24:24] like speaking in advance and and [24:26] starting the user group. So I love that. [24:28] Um I want to go uh talk a little bit [24:30] about about those things a little more [24:32] but just before I want to say hanoj uh [24:35] hey Fred hope you're doing well finally [24:37] here after a long time. Yes that was a [24:39] long time my friend I was worried about [24:42] you I was wondering what was happening [24:43] uh you didn't come for the last couple [24:45] of episodes but uh happy to see you here [24:48] always nice to have some friends on the [24:51] show and uh yeah hopefully uh you're [24:53] having a great day. So um on that note [24:56] uh Deepa u again you said so many things [24:58] that I would like to commend or hat [25:01] stuff on it. Um [25:04] first I like the fact that like you know [25:06] you're running a meetup. used to do a [25:08] lot of meetups. time I do damn I I I [25:10] don't run meetups anymore just because [25:13] um I do so many of those things part of [25:15] my day-to-day work that now on my free [25:18] time I'm more like you know what I don't [25:20] really want to do meetups or or speak at [25:23] events in my free time because I do that [25:24] for I love it but I do that for work and [25:26] I have other passions but um it's always [25:29] something that you can do like as an [25:31] individual if you're looking to you know [25:33] build some expertise if you want to go [25:35] within Devril if you want to connect [25:37] with the community if you to give back [25:39] to the community. Uh create a meetup and [25:42] depending where you live, maybe there's [25:43] not a lot of opportunity because there's [25:45] some cities where like there's so many [25:46] meetups that like there's no space for [25:48] more. But at many places, uh there's [25:51] probably like a niche topic where like [25:53] there's enough people in the community [25:54] that would be interested about that or a [25:56] more generic one depending where you [25:58] live and and creating a meetup. Yes, [26:00] it's it's work because you need to find [26:01] venues, you need to find speakers, you [26:03] need to do promotion, but I think it's [26:06] rewarding work and I love what you're [26:08] doing and I'm going there first to [26:11] encourage people to create meetups if [26:13] you have ideas. Uh it's it's it's work. [26:16] It's a little bit I would say easier [26:18] than people think. Meaning that people [26:20] like there's a lot of work, but like [26:22] once you start doing it, you have [26:24] hopefully like some other people is [26:26] going to be able to help you. some local [26:28] company who may help you to do venues. [26:30] You don't need to do them in person [26:31] also. You can do them online which make [26:33] it a little bit easier in terms of venue [26:35] or or finding uh those activities too. [26:38] But what I like one of the things you've [26:40] done and I didn't see that often is that [26:42] kind of collaboration you have with [26:44] other meetups uh where you just like [26:46] organize things together. So that helps [26:49] both organizers because it's one event [26:51] instead of two. uh that visibility to [26:54] different topics that that helps at so [26:56] many levels. So I just wanted to to I [26:59] guess I don't really have a question. [27:00] It's more me blabbering for you to like [27:03] to tell you. It's great idea. It's [27:04] great. It's I love that you're doing [27:06] that. [27:06] >> This this came up only last year, Fred. [27:09] It was not it was the new thing at [27:10] Google. So we were all expecting we were [27:13] initially creating external links to one [27:16] chapter. So what you would do is [27:17] manually go take if I if somebody is [27:20] hosting an event in GD GDG New York, [27:24] right? And I like that it's a virtual [27:26] event. I I personally want my chapter [27:28] members to know about this topic, right? [27:30] So I used to create an external ticket [27:32] and then it'll it'll reroute to their [27:34] page. It was a big hassle. Then then we [27:37] were all saying that is there a way that [27:38] we could just click one button and get [27:40] it get it done, you know, and now we [27:43] don't even have to do it. If you already [27:45] open to a co-host uh you know category [27:48] uh so if it's a co-host uh option is [27:52] anybody enables it it it relays it to [27:55] every other chapter so it's not one [27:58] particular um area and having said that [28:01] if people want to experiment I would [28:03] strongly suggest try to do uh an online [28:06] one see how you can build your community [28:09] first so uh it could be you can join a [28:12] GDG but if you want to do it, you know, [28:15] uh solo or start your own along or along [28:18] with your friends. Do a uh online first [28:21] and see how it how it goes because I [28:24] started during the pandemic. So I had to [28:26] do only online events to begin with [28:29] and the first year yes the first uh no I [28:33] did not get much traction. It was like [28:35] 100 or 200 people think slowly it grew. [28:39] So once I hit the mark of you know yes [28:42] we did a one uh you know uh in-person [28:45] event that's when people like in person [28:48] for sure but then the logistics trying [28:51] to create an event space and Google does [28:53] have some you know does have some um uh [28:57] guidelines or they have their own um I [29:01] would say there are some logistics to it [29:03] so now we don't do anything at the [29:06] Google offices anymore Initially earlier [29:10] we used to do it now they they I know [29:12] they want us to do it outside of the [29:14] campus which is totally fine but still [29:16] we have to follow their guidelines [29:18] making sure that everybody's over 18 and [29:21] if anybody is hosting in at the office [29:24] they have to bring their ID so uh in a [29:27] if you're doing it on the site you don't [29:29] need an ID as long as they are 18 years [29:31] old that's that's fine and even if [29:33] minors are coming they need to make sure [29:35] that these days kids also are interested [29:38] however ever um they have to come with [29:40] their parents. So that's another thing [29:42] and [29:44] >> I would nothing happens overnight. You [29:47] have to plan it accordingly. But when [29:48] you have a system in place, I have a [29:50] document. I just have to go it's now I [29:52] can do it uh you know for my sleep too. [29:55] It's not a big it's it just takes some [29:58] time to do it like how Fred you're doing [29:59] it with Streamyard and you know uh doing [30:02] it in Twitter all those things. But one [30:05] thing is uh I'm not as proficient as red [30:08] to streaming to different platforms. Uh [30:11] so we have our own platform at uh [30:14] Google. So that's how I do it. And you [30:18] can also live stream it on YouTube [30:20] directly there. Um I do have a GDG San [30:23] Jose YouTube channel too. But it's it's [30:27] initially I used to download everything [30:30] and then upload it. Now I don't do that. [30:33] So it's just a button away. So they made [30:36] it easier too. So it's this is uh this [30:40] is something that if you are starting [30:43] something on your own see whether Google [30:46] does require certain specific timelines [30:48] of doing the events. So you need to be [30:50] able to contribute to the community at [30:53] least tw you know once in twice two [30:56] months. So if you can um you know if you [30:59] have the time if you can do something [31:02] you can join the Google developer groups [31:03] near you or you can go you can start [31:06] your own you just have to contact and [31:08] then they will definitely let you know [31:10] and that's another thing another part of [31:13] being in the program is also um as a [31:16] organizer Fred there are when I say [31:19] campaigns they we also have free cloud [31:23] um certification campaigns too. So what [31:26] it would do is people can come in and [31:29] you get credits for that. It's not that [31:32] there might be a couple of raffle for [31:34] taking the exams for free but that is [31:37] not but you get all help from uh one of [31:40] the cloud experts to go through the [31:42] certification. So that's something uh [31:45] valuable if anybody wants to just look [31:48] for you know GDG [31:51] um any GDG that is organizing a cloud um [31:55] certification program it could be online [31:57] it could be an inperson if it is closer [31:59] to you go you know go uh take advantage [32:03] of that and learn the skills so the [32:05] other thing is any now the build with AI [32:08] campaign is coming and then they do give [32:10] you a Google cloud credit for you to try [32:13] out that particular workshop and that's [32:16] that's again uh you know something that [32:19] you can try it on your own that if you [32:21] you don't have to sign up for with your [32:23] credit card or anything because Google [32:25] does give you some credits for using the [32:28] tax AI and other stuff. So that's that's [32:32] another thing. Having said that, um you [32:36] want to do something on your own, [32:37] starting on on your own. Um start up [32:40] with a small meetup group online or your [32:44] local community and you build it over [32:47] over the over you know course of time [32:50] and [32:52] >> you'll connect with a lot of people. The [32:54] reason why I enjoy being at uh the [32:56] Google developer groups is because I'm [32:59] closer to u the community and the [33:03] professionals and also the technology. [33:06] So there are three parts to it and of [33:09] course being an expert also helps a lot [33:12] getting to know the product more and [33:14] there you know if you want to grow in [33:16] that area you could also do that. There [33:18] are different programs a IML they have [33:20] and Android they have flatter they have [33:22] even cloud and different technologies [33:26] just go the know go check with the [33:28] experts but you need to be invited for [33:30] the experts program [33:32] >> um by somebody and they would ask you I [33:35] do get a lot of people who who come and [33:37] ask me hey how do I become an expert I [33:40] said please read this blog first if you [33:42] have any questions please come back to [33:43] me because there's a it's it's it's it [33:46] you need to do the pre work before you [33:50] even apply to it. Somebody has to refer [33:51] you then they should be able to see [33:53] whether and please do not do these um AI [33:57] generated videos on YouTube and say that [33:59] I've created a channel I have different [34:01] followers. No, that's not authentic. [34:03] They want to see your face. They want to [34:04] they want to see how authentic you are [34:07] >> and that matters a lot. Even if you are [34:11] you don't need to be great at anything [34:13] but try try something. I know if you go [34:15] to my videos, I'm not proficient either. [34:17] I I still, but I still want to do it [34:20] because, you know, I someday I'll get [34:24] there like Fred, but I'm still trying. [34:27] >> So, no. [34:28] >> Yeah. [34:29] >> Awesome. Love to have. [34:30] >> No, I need to I need to learn from you [34:33] too, right? So, I I like to see how [34:35] people are doing and we all learn, [34:37] right? [34:37] >> We all need to learn from other people. [34:39] So, no, there there's a lot of bit a lot [34:40] of benefits to those programs. It's a [34:42] great recognition, but there's benefits [34:44] to and it's fine if you're not part of [34:46] one too. I mean like because they're [34:47] there limited spot. Um I guess also my [34:50] my question my point was like yeah if [34:52] you can build something if you can [34:54] create stuff fine but also if you don't [34:56] have the time you don't have the energy [34:57] you're not sure where to start uh just [34:59] go to those things. to go to meet tops, [35:01] go to the events, you meet people, you [35:03] create connection, you don't know the [35:04] person you met at an event, if there [35:06] will be potential uh someone who's going [35:08] to hire in their team in the future in [35:10] your future bus, future colleagues or [35:12] whoever. And it's just just a great [35:14] opportunity. So there is more to those [35:16] meetups and events than like the the [35:18] technical content by the speaker or [35:20] whatever the content the speaker is [35:22] sharing. So with that note, um we could [35:25] talk forever. Uh I have so many other [35:27] things I would like to discuss with you [35:29] but uh I think it's time for us to move [35:32] to uh the second part of the show which [35:34] is the uh demo and um I have a huge [35:38] memory blank right now. Uh I forget like [35:42] I I know it was something about AI but [35:44] I'm so sorry it's totally unprofessional [35:46] but like I really don't remember what [35:48] was the deal. [35:49] >> No worries. So it was multi- aent. I I [35:52] saw that there was a lot of things with [35:54] uh you know lang chain and a different [35:57] backend base. So I thought okay maybe [35:59] let's let's do something different um [36:01] and do if if you have to be a you know [36:06] uh been if you have to be in in a full [36:08] stack environment and you want to code a [36:11] code review pipeline how would you do [36:13] that? It's just mimicking um in a real [36:16] life scenario. So I'm going to be let's [36:18] see how I can share here. Uh I'm gonna [36:22] Okay, I'm not going to be seeing you, [36:23] but uh I'm going to minimize this thing. [36:26] >> That's fine. You're not missing so that [36:28] >> Okay. So I'm going to sorry. So this is [36:32] what we going to be doing, right? So if [36:35] you see here in a real life scenario, [36:37] you will be just again know the [36:39] developer goes pushes and then there is [36:41] a VR open CI is triggered, right? And [36:44] then what this demo is going to be [36:47] teaching know I'm going to be talking [36:49] about is how do you add a AI agent to [36:52] the pipeline. So you can add three [36:53] agents. I chose chat you you could use [36:57] uh you know cloud for everything but I [36:59] chose to show two three I know two [37:02] different um um um AI providers here. uh [37:06] one is uh you know the reader agent from [37:09] open AAI because it's good for [37:11] summarizing and uh summarizing and I'm [37:15] using two agents a reviewer agent and [37:17] the fixer agent. So in a real life [37:20] scenario the agent would go check for [37:24] your code. I'm going to show this in a [37:25] real quick moment. Before I do this let [37:28] me uh I have to see what I did this. [37:32] Okay. So, where is my thing? So, [37:37] bear forward. Can you see my code? [37:39] >> I can see your code. [37:41] >> Okay. Cool. Uh, so let's start with the [37:44] rout. It's I'm not going to be going too [37:46] much into it. You can everybody can go [37:49] do this. It's very super easy. But I [37:51] wanted to show that we can do it for [37:52] backend and front end scenarios. So, [37:55] this is the route agent where I have it [37:58] for backend and as well as for front [38:00] end. And there is a reader agent where [38:03] I'm using OpenAI [38:05] uh GPT model for uh O mini and then it [38:09] reads so what it it it can summarize it. [38:12] It just summarizes what the code has [38:14] given right and then the second one is [38:16] uh the reviewer which is from cloud it's [38:18] using cloud sonnet and it just uh [38:21] reviews the entire code and sees what [38:23] are the potential issues here. Third uh [38:27] one is a reviewer one which is also from [38:30] oh sorry the fixer one which is also [38:33] from clot and here it just it does not [38:37] fix your code but it does say that hey [38:39] there are some issues. This still would [38:41] need a human eye to see whether you can [38:45] push it to the pipeline or whether you [38:48] can u you know merge it or not. So this [38:52] is some and I created couple of [38:54] scenarios I think seven of them I think. [38:57] So uh let's see the backend scenario um [39:00] see these are the agents that I'm using [39:02] uh reader agent and then the reviewer [39:05] agent and the fixer agent and the [39:07] backend scenarios how do you do if you [39:09] do a hardcore this I have seen it very [39:11] commonly so if you're running tests or [39:13] something people do run this uh and your [39:17] information is getting you know exposed [39:19] so in that case howard uh claude claude [39:22] and uh GPD catches it right so that's [39:25] that's another thing and API key is [39:29] exposed. We I do this very very often [39:32] when I'm doing the demo one. I usually [39:35] when I try to use uh uh you know when [39:38] I'm using uh you know I forget so I [39:42] forget to create it uh when I'm testing [39:45] it I just directly enter the API key and [39:48] then I expose it to so then I realize oh [39:52] my god I forgot to create a [39:54] environmental file right so this happens [39:56] quite often and this is um this would [40:00] also the backend one also tries to [40:02] create a mimicking uh SQL injection [40:05] vulnerability and then the login one and [40:09] also a ne for performance um based if [40:12] it's a nested loop it's going to show [40:14] that hey there are three uh this is [40:16] going to be uh work n3 so this is a [40:19] performance disaster so this is going to [40:21] show and then on the front end side um [40:24] it is going to be um selecting a card [40:27] and then how is the how is it uh missing [40:31] the there is a missing um um alt text [40:35] and then there is this I have seen you [40:38] know the the layout changes for example [40:42] if you're using RTL languages um to the [40:45] Latin one so how does it fix it so this [40:48] is what there are some scenarios and the [40:50] color and stuff like that so I'm going [40:52] to be running this demo real quick if [40:54] you have quick time in here and then you [40:56] can take a look at this also and let's [41:00] see how we can run this so So I'm going [41:02] to run the first the reader agent for [41:04] the back end. So for example I'm running [41:07] it hopefully it is running. So it does [41:09] say okay it does summarizes it saying [41:11] that hey this is the typescript one. So [41:14] it is connecting to the database and [41:16] then we are connecting to the middle [41:19] layer and then the API endpoints are so [41:21] this is done with uh chat GPT uh the GPT [41:25] uh for uh model and then the reviewer [41:28] agent it looks for okay there are some [41:31] issues. So there is a SQL um thing [41:35] critical issues and it also says that [41:37] there is a hard-coded secret and once [41:40] it's all done it just it gives you the [41:42] fixer one does not fix anything but then [41:45] it says that you could you could look [41:47] through it. This still requires a human [41:51] eyes. So make sure that whether these [41:54] issues that are uh reviewed by the [41:57] reviewer agent uh somebody could go and [41:59] fix it. So it would fix it but still [42:01] needs somebody's eyes in there [42:04] >> right. So this is one part. This is [42:07] done. And then what I found okay this is [42:10] not relatable. It had hallucination. [42:12] Then uh no I I tried to give it specific [42:15] uh examples here. Uh it was [42:17] hallucinating. Then I had to it is not [42:20] right the it was not right the first [42:22] time when it uh the reviewer agent had [42:24] issues. it was hallucinating when I was [42:26] using anthropic the first couple of [42:28] times and I said hey how is this you [42:30] know you're saying that it's a SQL [42:33] injection vulnerability but then it's [42:35] just testing it you know using for it it [42:38] it had a issue in here and let's go to [42:42] the front end one so what it it give you [42:45] um I I made it in such a way that it can [42:48] show the visual part of it too so again [42:51] it tries to create look for the [42:56] um product card with the with details [43:00] and if you see there are some I [43:02] personally asked for if it was using a [43:05] mobile how the layout changes and what [43:09] are the what are some of the things it [43:11] does say that there are there are no [43:12] critical issues but there are some [43:14] issues here [43:16] just give me one second while it [43:18] processes it and then it'll show exactly [43:22] and uh What one thing I found out was if [43:26] you have done some um [43:29] uh internationalization there are some [43:31] ways that the numbers are displayed in [43:34] different ways. Hang on. Let me see. It [43:36] was working last night. Trust me, it was [43:40] let's give it a couple of [43:43] go. So yeah, this so if you see here um [43:47] it checks for the vet and and the [43:50] pricing, right? This is in English. Uh [43:53] it is it is okay. It is it is like this [43:57] the beginning and then it got fixed. So [44:00] again the layout itself has uh has been [44:03] fixed too. So if you see Arabic the [44:06] initial one you see the the number the [44:09] currency has not changed here and [44:12] depending on the browser language [44:15] setting because I was when I was testing [44:18] this whether this you know the numbers [44:19] were not like this right it was giving [44:22] something in Arabic so I don't know [44:24] Arabic but then I saw that hey why is [44:26] this numbers or not numbers I could see [44:29] a number and a text difference right so [44:32] depending on it did not Even cloud did [44:35] not catch it. So I said, "Hey, you know, [44:37] this is something wrong." So don't trust [44:40] with what the agent says. You have to [44:44] know whether it's doing the right thing [44:45] or not. That's why I said that you still [44:47] need a the fixer agent would not fix [44:49] anything. You still have a you need a [44:51] human intervention here. So like I said, [44:54] then the text have been changed and also [44:59] oh the the German one here, Germans [45:02] write the currency in a different way. [45:03] So the N is at the the euros at the end [45:05] and if you see there's a comma in here [45:08] and little and everything is also um in [45:12] German here earlier it was not. So these [45:14] are you know the common things in uh [45:18] front end um development you would use [45:21] and then the back end backend one of [45:22] course you can try to play with it you [45:24] can this is an agnostic one I just chose [45:28] um uh no open AI and anthropic however [45:32] you can use it you can use u anthropic [45:36] for for all three however um open AI's [45:40] um um model is cheaper and it it is [45:43] faster faster for in terms of [45:45] summarization. So that's why I want to [45:46] just choose this. If you are working at [45:48] a place where you know if you're working [45:50] at a company and they can afford then [45:52] you can use all three as cloud or you [45:54] can use all three as open too. So that [45:58] would be my demo. [46:01] >> Yeah, [46:01] >> hang on. So that was quick. If you want [46:03] to take a look at it, it should be in my [46:06] in my GitHub [46:08] >> wonderful. If it is not there, I will [46:10] push it because I forgot to I will make [46:12] sure that my API keys are not exposed. [46:14] You this is a paid version. So, uh it's [46:16] it's not that much. You just have to pay [46:18] $5 and so that you can experiment it. [46:20] I've been experimenting. [46:21] >> Yeah. Yeah. Still still we don't want [46:23] your uh your your API key exposed. No, I [46:25] like that because uh there's uh things [46:28] that happen with AI and and we're going [46:30] to close on that. Uh but you know like [46:33] there's different cost. It cost a lot of [46:35] money if you if you use it a lot. So [46:37] like trying to optimize for different [46:39] models makes sense. So uh it's like I I [46:42] feel hope like it's like when we're [46:44] talking about cloud you know like when [46:46] cloud would becoming a thing and we're [46:47] like no don't just put all your eggs on [46:49] the same cloud like you use multi cloud [46:51] approach because they have like like [46:53] it's a little less true right now but [46:55] like before each had their own pros and [46:57] cons and like you could use like a [47:00] specific cloud for specific task and [47:01] another one for another task. And [47:02] anyway, it was just a great way to be [47:04] sure you don't have like a single uh [47:07] what do you call that like a single [47:08] entry failing entry point something that [47:11] so I guess that's the same thing too [47:13] here and it's optimizing like the cost [47:16] and and using the best model for what [47:18] the task you have to do [47:20] >> you could also even prior to using this [47:23] model you can also use something because [47:24] there are some static ones called a [47:26] sonar cube or something uh like that you [47:28] can also uh use that along with the [47:30] pipeline so that you don't have to you [47:32] know use so much of your tokens too. [47:35] >> Yeah. Yeah. And and I love also that you [47:38] said that I have funny enough I'm [47:39] speaking on the meet tops tonight about [47:40] AI and uh one of my big obvious takeaway [47:45] of my talk is the like there's some [47:48] benefits of AI but there's also some [47:49] risk. There's also things that that [47:52] people don't talk about enough that we [47:54] should talk about. But uh one of the big [47:56] thing that I want people to come back is [47:57] like like never remove the human from [47:59] the process especially on right hand. [48:02] Maybe in the future we'll see but right [48:03] now Alice Nation uh sometimes the LLM [48:07] are trained on old data. Uh you cannot [48:10] like they're doing a pretty good job but [48:11] you cannot like bly trust any agents [48:14] right now. So always have a human being [48:16] part of the process to be sure that [48:18] what's fixed, what proposition, what's [48:20] change makes sense. Uh I think I think [48:22] it's really important. So it may not [48:23] seems like you're going to be as [48:24] productive is like if you automate [48:26] everything, but at the end of the day, I [48:28] think you're going to end up with better [48:29] quality and it's going to be better in [48:31] the medium long term. So [48:33] >> and also like I said, this could be [48:35] agnostic, right? So you can you can swap [48:37] it with llama or you can swap it with [48:39] Gemini. But then I did try for with [48:41] Gemini too. I I I ping you. I told you [48:43] to. So if you uh I was trying the Gemini [48:46] 1 no 2 and 1.52 I was trying testing it [48:50] was not working the reading reader agent [48:52] was like okay maybe I and you need to [48:56] use vortex AI for that I tried with one [48:59] of my free cred and the credentials that [49:01] I had. So if you have the time and I I [49:04] would say that if you're familiar with [49:06] GCP and Vertex AI then go with it. I [49:09] wanted to show a demo that if somebody [49:12] wants to show it, if somebody wanted to [49:13] use it, a very very basic uh plugand [49:16] play, this is what this would be easier. [49:19] >> That's awesome. So, uh thanks for the [49:21] demo. Thanks for uh sharing your your [49:24] passion. Uh I think people saw it. [49:26] You're really passionate. You're really [49:27] excited about things which I love. Uh [49:29] that also makes my life easier as a MC [49:32] because I was like I was asking you a [49:34] question and you were going and like [49:36] like like talking about it with passion [49:38] and everything and it was that's [49:39] amazing. Uh I'm not even joking. That's [49:41] great. I'm not even teasing you. I love [49:43] it. Uh the teasing part is that there's [49:46] rarely people that talk more than me and [49:48] you're one of those persons. So uh you [49:50] decide if it's a good or a bad thing. I [49:52] like it because I talked a lot. But uh I [49:55] mean that was that again uh thanks for [49:58] being in the show. That was amazing. Is [50:00] there any last thing you want to leave [50:02] people with before we uh close that [50:04] episode? [50:06] >> Um well I just started a newsletter. Uh [50:09] so this is also one way that you could [50:11] start with the community if you want to. [50:12] I started with a newsletter and if you I [50:15] I am I'm going to be focusing more on AI [50:18] platform and architecture on my [50:19] newsletters. Um so if you want to sign [50:21] up it's there on my LinkedIn. I just [50:24] started this uh so yeah and I would also [50:27] be running few uh from March 5th I'm [50:30] planning to run a forum just come in [50:33] it's going to be free it's going to be [50:35] more of AI architecture platform all [50:37] those stuff so I'm you know it's not [50:39] it's not hands-on work I'm not going to [50:42] do but then we can discuss like how we [50:44] talked about today it's bug we we talked [50:46] about this we could also talk about how [50:48] what are the things that breaks or how [50:50] is the cost going to implement you know [50:52] cost going uh [50:55] uh how it's cost effective or what [50:57] whether should I use it whether should I [50:59] not use it all those stuff and how do we [51:01] architect and what are the mistakes that [51:03] we go along the way we can talk all [51:05] about it too [51:07] >> that sounds good so you're a busy woman [51:09] so uh thanks again for being on the show [51:11] was a pleasure having you uh thanks [51:13] everyone who joined us live thanks to [51:15] the people also who's going to watch the [51:17] recording and uh I'm not sure if we're [51:20] going to see each other in two weeks [51:21] because uh I think In two weeks I'm [51:23] going to be in Europe. I'm going to be [51:25] in Amsterdam for Laracon EU. Also [51:28] speaking at a ride the docks meetups [51:31] there and you know getting busy. So I [51:34] won't have an episode in two weeks but I [51:36] think and I do not know my schedule but [51:38] I think I'm going to be back in about a [51:40] month with another episode. So thanks [51:42] everyone. Thanks Diva. Have a good rest [51:45] of the day and see you soon folks.  ## 📺 Video ![](https://www.youtube.com/watch?v=fXJW_zEeZco) --- # 📝 Full Notes (Deep-Read Analysis) # 🏷️ Title: DevRel Show Episode 45 - Community Leadership and AI Code Review with Deepa Subramanian ## 🧰 Action Inbox - [H1] Apply to Google Developer Expert program - requires invitation and community contribution #gde - [H2] Start local meetup or join existing GDG chapter - build community connections #community - [H3] Implement multi-agent AI code review pipeline - use different models for cost optimization #ai ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ### 🔗 References - [DevRel Show](https://devrel.show) - [GDG San Jose](https://gdg.community.dev/gdg-san-jose/) - [Deepa's LinkedIn](https://linkedin.com/in/deepa-subramanian) ## 📝 Summary - Episode 45 of DevRel Show featuring Deepa Subramanian discussing community organizing and AI-powered code review - Deepa shares her journey from non-traditional tech background to Google Developer Expert and GDG organizer - Demonstrates multi-agent AI system for automated code review using OpenAI and Anthropic models - Emphasizes importance of human oversight in AI processes and community building through consistent effort - Key takeaway: Community growth requires patience, collaboration, and authentic contribution over time ## 📖 Main Content Deepa Subramanian represents the growing trend of non-traditional tech professionals becoming community leaders. Her transition from academia (PhD background) to tech demonstrates that diverse backgrounds bring valuable perspectives to developer relations and community organizing. The Google Developer Expert program requires invitation-based application with four interview stages: screening, community assessment, product interview, and Googler interview. Success depends on authentic community contribution rather than artificial metrics like AI-generated content. Her AI code review demo showcases practical applications of multi-model AI systems, using OpenAI GPT-4 for summarization and Anthropic Claude for detailed code review and fixing suggestions. The system identifies security vulnerabilities, performance issues, and accessibility problems while maintaining human oversight as a critical component. ## 🗒️ Notes #### ⏱️ [00:00](https://youtu.be/fXJW_zEeZco?t=0) — [H1] Episode Introduction and Technical Difficulties - Key idea: Show rescheduled from 3pm to 2pm Eastern, causing StreamYard complications - Evidence: "I scheduled it at 2pm but I did the promotion for 3pm" #### ⏱️ [02:30](https://youtu.be/fXJW_zEeZco?t=150) — [H2] Deepa's Background and Community Journey - Key idea: 15+ years industry experience with non-traditional path to tech - Evidence: "I come from a non-traditional route... I was a woman techmakers ambassador" #### ⏱️ [05:15](https://youtu.be/fXJW_zEeZco?t=315) — [H2] Starting GDG San Jose Community - Key idea: Started from scratch in 2021, now has 1,400+ members - Evidence: "Community is not built in a day. It takes consistent effort" #### ⏱️ [08:45](https://youtu.be/fXJW_zEeZco?t=525) — [H3] Teaching and Mentorship Motivation - Key idea: Personal experience lacking mentors drives desire to help others - Evidence: "I never had a mentor... if at all I could be somebody's mentor" #### ⏱️ [12:20](https://youtu.be/fXJW_zEeZco?t=740) — [H1] Google Developer Expert Program Details - Key idea: Four-stage application process requiring invitation and authentic community contribution - Evidence: "You need to pass the screening... four or five steps" #### ⏱️ [18:30](https://youtu.be/fXJW_zEeZco?t=1110) — [H2] Community Event Collaboration - Key idea: Co-hosting events with other chapters increases reach and reduces organizer burden - Evidence: "If anybody enables co-host option, it relays to every other chapter" #### ⏱️ [22:15](https://youtu.be/fXJW_zEeZco?t=1335) — [H3] Paper-Based Learning Preference - Key idea: Visual learning through handwritten notes improves comprehension - Evidence: "I still do write in paper and pencil... I draw stuff" #### ⏱️ [25:40](https://youtu.be/fXJW_zEeZco?t=1540) — [H2] Benefits of Expert Programs - Key idea: Programs provide early access to technology and direct product team feedback - Evidence: "You have access to information prior... you can suggest these things" #### ⏱️ [32:10](https://youtu.be/fXJW_zEeZco?t=1930) — [H1] AI Code Review Demo Introduction - Key idea: Multi-agent system mimics real-world CI/CD pipeline with AI integration - Evidence: "How do you add AI agent to the pipeline... three agents" #### ⏱️ [35:45](https://youtu.be/fXJW_zEeZco?t=2145) — [H2] Multi-Model AI Strategy - Key idea: Different AI models optimized for specific tasks (OpenAI for summarization, Claude for review) - Evidence: "OpenAI's model is cheaper and faster for summarization" #### ⏱️ [38:20](https://youtu.be/fXJW_zEeZco?t=2300) — [H2] Backend Security Issues Detection - Key idea: AI agents identify SQL injection vulnerabilities and exposed API keys - Evidence: "Hard-coded secret... SQL injection vulnerability" #### ⏱️ [42:30](https://youtu.be/fXJW_zEeZco?t=2550) — [H3] Frontend Accessibility and Internationalization - Key idea: AI reviews detect missing alt text and RTL language layout issues - Evidence: "Missing alt text... layout changes for RTL languages" #### ⏱️ [45:15](https://youtu.be/fXJW_zEeZco?t=2715) — [H1] Human Oversight Requirement - Key idea: AI hallucination and errors require human validation in all processes - Evidence: "Still requires human eyes... don't trust what the agent says" #### ⏱️ [48:30](https://youtu.be/fXJW_zEeZco?t=2910) — [H3] Newsletter and Future Plans - Key idea: Starting newsletter focused on AI platform architecture discussions - Evidence: "Just started a newsletter... focusing on AI platform and architecture" ## 📚 References - [GDG San Jose YouTube Channel](https://youtube.com/@gdgsanjose) - [Google Developer Expert Program](https://developers.google.com/community/experts) - [Women Techmakers](https://womentechmakers.com) - [Deepa's Medium Blog](https://medium.com/@deepa-subramanian) ## 🗣️ Paraphrase - Community building requires consistent effort over time, starting small and gradually growing through authentic engagement rather than expecting immediate large audiences - The Google Developer Expert program values genuine community contribution and technical expertise over artificial metrics, requiring invitation and multi-stage interviews - AI-powered code review systems should use multiple specialized models for different tasks while maintaining human oversight to prevent hallucination-related errors ## 📰 Newsletter 🚀 **Community Leadership Meets AI Innovation** This week's DevRel Show highlighted how non-traditional backgrounds can bring fresh perspectives to developer relations. Deepa Subramanian's journey from academia to Google Developer Expert demonstrates that authentic community contribution matters more than formal credentials. **Key Takeaway**: Multi-agent AI systems for code review show promise, but human oversight remains critical. Consider using different AI models for specialized tasks - OpenAI for summarization, Claude for detailed analysis - to optimize both cost and accuracy. **Practical Application**: Start small with community organizing. Online-first meetups reduce logistical complexity while building initial audience. Collaborate with existing communities through co-hosting to expand reach efficiently. **Resources**: [Learn WordPress](https://learn.wordpress.org/), [Developer Docs](https://developer.wordpress.org/), [WP-CLI](https://wp-cli.org/) ## ✅ Action Items #### DevRel - Developer advocacy: Create authentic content demonstrating real expertise rather than AI-generated materials - Education: Develop mentorship programs for non-traditional tech backgrounds entering developer relations - Developer experience: Implement multi-model AI systems for code review while maintaining human validation - Community management: Establish co-hosting partnerships with adjacent communities for expanded reach #### Community - Maintainers: Document multi-stage application processes for recognition programs - Contributors: Focus on consistent, authentic contribution over artificial engagement metrics - Ecosystem: Develop collaboration frameworks for cross-community event organization ## 🧪 Hypotheses - Multi-model AI systems provide better cost-performance ratios than single-model approaches for complex code review tasks - Community organizers from non-traditional backgrounds bring unique perspectives that enhance developer engagement - Co-hosted virtual events generate higher attendance and engagement than single-community events - Paper-based learning techniques improve technical concept retention compared to digital-only methods ### 📊 Methodology - A/B test single-model vs multi-model AI code review systems measuring accuracy and cost - Survey community members on organizer background diversity impact on event quality and inclusion - Compare attendance and engagement metrics between solo and co-hosted virtual events ### ❓ Questions - How can recognition programs like GDE scale while maintaining quality standards and authentic contribution requirements? - What combination of AI models provides optimal cost-effectiveness for different code review tasks? - How do non-traditional career paths contribute unique value to developer relations roles? ### 🔮 Future Research - Impact of human oversight percentage on AI code review accuracy and team productivity - Correlation between community organizer diversity and long-term member retention rates - Effectiveness of paper-based vs digital learning methods for technical skill development ### 💡 Implications - Practice: Implement multi-stage validation processes for AI-assisted code review - Policy: Develop guidelines requiring human oversight in AI-powered development workflows - Education: Create pathways for non-traditional backgrounds entering developer relations - Funding: Budget for multiple AI model subscriptions rather than single-provider solutions ### 🧠 Learning Styles Assessment Reading level estimate: 12th grade #### 👀 Visual learners - Present: Code examples, architecture diagrams in demo - Improve: Add flowcharts showing GDE application process, community growth metrics #### 🔊 Aural learners - Present: Conversational interview format, verbal explanations - Improve: Provide audio-only version with enhanced sound quality #### 🗣️ Verbal learners - Present: Detailed explanations, analogies, technical terminology definitions - Improve: Add glossary of community organizing terms, more real-world examples #### 🧩 Kinesthetic learners - Present: Live code demo, step-by-step technical walkthrough - Improve: Provide downloadable code samples, interactive setup guides #### 🧠 Logical learners - Present: Four-stage GDE process, systematic approach to AI model selection - Improve: Add decision trees for model selection, ROI calculations for community events #### 🧑‍🤝‍🧑 Social learners - Present: Community collaboration examples, co-hosting strategies - Improve: Include discussion prompts, community building checklists #### 🧘‍♀️ Solitary learners - Present: Individual learning journey narrative, self-directed skill development - Improve: Add reflection questions, personal development frameworks #### 🌿 Naturalistic learners - Present: Organic community growth analogies, ecosystem thinking - Improve: Connect to broader tech ecosystem patterns, sustainable growth models ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Conversational format with clear speaker identification - Improve: Add detailed alt text for code demonstrations, structured headings #### Keyboard navigation - Present: Standard video player controls - Improve: Provide transcript with timestamp navigation links #### Contrast and readability - Present: Clear video quality, readable code samples - Improve: Increase code font size, ensure high contrast in IDE themes #### Media and captions - Present: Full transcript available, clear audio - Improve: Add professional captions with speaker identification, code snippet descriptions ### 🌍 DEIB Assessment #### Cultural breadth - Present: Global community perspective, international collaboration examples - Improve: Include more diverse geographic examples, multilingual community strategies #### Inclusive language - Present: Gender-neutral terms, welcoming community language - Improve: Define technical jargon consistently, avoid cultural idioms #### Neurodiversity support - Present: Clear structure, consistent pacing - Improve: Provide content warnings for rapid technical demonstrations, offer alternative formats #### Representation - Present: Non-traditional career path representation, diverse professional backgrounds - Improve: Include more varied identity representation in examples, accessibility considerations ### ✅ Logical Validity #### Strong reasoning - Deductive strength: GDE program requirements → community contribution necessity → authentic engagement focus - Inductive strength: Individual community success → broader community organizing principles - Causal reasoning: Consistent effort → community growth → sustained engagement - Analogical strength: Academic mentorship needs → tech community mentorship needs #### Fallacies - Survivorship bias: Success story focus without failure case analysis - could include balanced perspectives on challenges - Appeal to authority: GDE/Google credibility emphasis - balanced with independent validation of strategies #### Tone of voice - Overall tone: Enthusiastic and encouraging with authentic vulnerability about challenges and mistakes ### 🛡️ DARVO Analysis - Deny: No denial patterns present - Attack: No attacking behavior observed - Reverse victim/offender: Not applicable to this content - Emotional cues and rhetoric: Positive, supportive language encouraging authentic participation and learning ## 🎯 Conclusion Community building and AI implementation both require patience, authenticity, and human oversight. Deepa's journey demonstrates that non-traditional backgrounds bring valuable perspectives to developer relations, while her AI demo shows the promise and limitations of current automated systems. Success in both domains comes from consistent effort, genuine contribution, and maintaining human judgment in automated processes. ## 🧱 Refactor Appendix ### 📚 Literature Split Pad Purpose: Stage CEE units that become Literature notes. Split level: H3 ### [H1] Google Developer Expert programs require authentic community contribution > [!quote] DevRel Show Episode 45 by [[Frederic Harper]] via [[YouTube]]. Published 2024. Accessed 2024-12-28. > "Please do not do AI generated videos on YouTube and say that I've created a channel. They want to see your face. They want to see how authentic you are." Authenticity over artificial metrics in recognition programs. [view highlight](https://youtu.be/fXJW_zEeZco#H1) #xref #articleseed ID:rw_gde_authentic ### [H2] Multi-model AI systems optimize cost and performance > [!quote] DevRel Show Episode 45 > "OpenAI's model is cheaper and faster for summarization. So that's why I chose this." Strategic model selection based on task-specific strengths. [view highlight](https://youtu.be/fXJW_zEeZco#H2) #xref ID:rw_ai_multimodel ### [H3] Community growth requires consistent long-term effort > [!quote] DevRel Show Episode 45 > "Community is not built in a day. It --- # 📄 Full Transcript *Auto-generated from YouTube captions · timestamps roughly every 30 seconds · [watch on YouTube](https://www.youtube.com/watch?v=fXJW_zEeZco)* **[00:31]** Hello, hello lovely people. Welcome to another episode of the Devro Show. I think it's episode number 45. And uh hopefully uh you're able to join us right now. Uh that was not my best management of an episode of the show. You know, usually the show goes at 3 p.m. Eastern time and I scheduled it at 2 p.m. But I I did the promotion for 3 p.m. And uh like because of that, I had to reschedu things in StreamYard and I had to delete some stuff on LinkedIn and **[01:02]** YouTube and recreate it. So hopefully you find the right link to uh join us live today. In uh worst case scenario, uh the lovely thing with those webinars is that there's always the recording available. So, you're going to be able to uh watch the show uh after if you're not able to join us live. But hopefully, it's always fun to have people live uh with questions and comments. Uh but you know, as I said, sometimes even when there are people live, I didn't tell you, but like sometimes we have people live and it's just like nobody's there because nobody's asking question, nobody's commenting. And sometime we get **[01:33]** 10,000 of those. So it really depends on like you know how's the sun today, what's the weather, what's the alignment of the stars. Like there's no real way to uh understand exactly when we have a lot more activities or not. But uh always happy to do the Devrell show especially since uh I started to travel again for work. They're a little less uh every other week like they used to be. So sometimes there's a little more time between each other. But uh I continue to do those. Still really excited. I have a really long list of uh guests in the **[02:04]** wait list. So uh they won't stop uh tomorrow. So again really happy to be here. It's always fun because I always have amazing guests. And today with me I have uh Deepa. How are you? >> Hey, thank you so much Brad. I'm doing very well and no worries about the whole no Streamyard YouTube matchup thing. You're totally fine. >> Wonderful. Wonderful. So thanks for joining. Uh we waited a long time. Uh sorry. >> So sorry. Uh do I have to join the chat here? Uh so **[02:35]** >> no, you're good. >> I'm good. Okay. Thank you. >> No, no, you're good. You're good. If there is anybody uh you know commenting, I will share them on the screen. Uh so yeah, no, you're good here. But uh no, thanks for uh thanks for being a guest on the show. Uh now that's probably the story. Unfortunately, everybody who wants to be uh on the show wants to be a guest uh because there's a long wait list and I don't do them as often as I would like to because of my schedule. Uh it takes a little bit of time. So, I think we talked about this a couple of months ago if a year ago, but finally I was able to have you on the show. So, **[03:07]** first question for you. Uh who are you? >> That's a good question. So, well, um for people I'm I'm not new to the community actually. Um, for people who know me, uh, I'm Deepa. I have been in the industry for 15 plus years, but I also run a lot of communities around my local area. I live in San Jose, California. I run GDG San Jose. And I there was a part that, um, before I even started out with San Jose, um, I was a woman techmakers **[03:38]** ambassador. Now it's been moved on to Technovision. Um so yeah uh been in around with the community for more than five years now. So professionally I've been working for more than 15 plus years. Uh I have come a long long way. So yeah my previous my last job was at Adobe and then um I my career path was totally a different one a non-traditional one as well for people who do not know me. I come. Yeah, it's **[04:09]** it's a totally different route that I took uh back in the days. So, but I'm excited and I'm also happy that I made that decision. So, yeah, that's me. >> No, that's great. >> If you want to follow and if you want to follow me everything, Fred will post it at the end. >> I will share a lot of link actually. Uh you just remembered me that I have a lot of links to share and I for whatever reason I was so into what you were saying that I forgot about those. But yeah, during the show I'm going to share a lot of links uh where uh places where you can connect with depot and uh really **[04:40]** happy uh actually I have a little bit memory so I don't know for how long we know each other. I know I've I've spoke once at your user group and it's also part of the reason why um I was happy to have you outside of the fact that you're an amazing person but you know like it's a dev show. So, usually the guests I have on the show, uh, they usually have the title developer advocate or developer relations something or technical evangelist or like the kind of title where like you your full time is doing that. But, uh, once in a while because I **[05:11]** don't want to restrict myself. Uh, I want also people that may not have like the official title but still do some or many of the things that developer advocates do, developer relation. And part of it is uh your involvement in the community is the fact that you're a community organizer as you said like the GDG uh San Jose and uh for me those are some of the things that usually depending on a company depending on your role uh you know uh developer advocates may do uh in in in their day-to-day job. **[05:43]** So I invited you today as like a community community organizer. So I would like to you know uh start with that uh you know >> why did you start a community? Why are you doing that? >> What are you doing? >> Um so the story goes behind in 2021. Um so um the thing was I started off with a whole community thing. Google gave me an opportunity to talk at um uh you know talk uh be in a program called as women developer academy back in 2021 and they **[06:14]** were training about 30 women for this academy and I was chosen it was there were 300 applicants and I was chosen one so uh amongst the 30 people so that's how I got a step or I came to even know about this community um community as a whole what I could contribute to the community I you know personally I'm very close to my I know teaching is something that is very close to my heart and tech is something that I just you know I chose a way to transition to and when I was uh **[06:48]** you know getting into tech I have had my own struggles to grow and I I never had a mentor so that's I was if if at all I could do and do something and if at all I could be somebody's mentor. Um that's that's how it all started. So I I do love speaking. Um this was not in technique uh tech part of my uh life but then I've I've done a lot of um when I was uh in my PhD program. I have given **[07:18]** talks, I've written research work, papers and all those other stuff. So public speaking was never a fear or anything like that. But how do I bring it to the you know bring bring it into tech was a was something that I did not know. When Google gave me this opportunity in the WBA program I stepped onto it right away. And then they said hey do you want to do events? Do you want to be speaking somewhere? I'm like oh I could I want to do this. So they said that you could teach you could **[07:49]** inspire people with what you know. So I was not very sure how people would take me at the beginning or would they even accept me from coming from a non-traditional uh I was a bit hesitant. So then once I started doing it I'm like okay this is fun. So and when I started uh speaking as well as this is when they said hey can you also do you know events? So I create I started this GDG San Jose in **[08:20]** California. um in 2021 I guess. So yeah and I started off from scratch. So for people out here your community is not built in a day. It takes consistent effort and today I'm very happy that there are about 1,400 members in the community that I run it. So it it does but last year was a little bit of um um little bit I was not very active but this year I I will be active so that's **[08:53]** something we are you know I'm planning to do a lot of events during this year and Google as a whole they you know they have campaigns for every quarter so um earlier it used to be about four events per year now the count has become uh six. So that's something that you know I have to schedule it accordingly and do it. And this this is how I started and when I say events I've done about I **[09:25]** don't have a count I think last year I think 60 to 70 events uh together but then I also co-host with other events too. So if anybody is doing across in the US or anywhere in the globe and they said hey do you want to co co-host with uh with my chapter. So I co-host with them too. So I'm invol I know that there are um couple of things one in Sudan and uh there are a couple of other chapters they ping me the organizers ping me and ask hey can you co-host it I'm open to but provided it's in the time zone and **[09:57]** the right time it's there. So um so you collaborate too that way. And the third part of the community was um I was while I was also speaking and giving talks um I became a Google developer expert too. So this is something uh yeah I did not say that to you Fred because >> last year yeah last year my uh and Google you know for any events that you do any talks that you do as a GDE you we **[10:30]** have a reporting system so uh I have to report it but then in in doing events is pretty easy once you're done you have to wrap up the event and say that hey this and it automatically uh you know it'll tell you how many if it's an online event it'll that tell you how many online attendees were there. Um, so the event wraps up and it automatically does it. However, uh, in the GD program, you have to go and, uh, probably put in all your details and that's where I forgot **[11:02]** last year. I would be very and I kind of lost it. Um and I had to talk to the G uh this and I it just got reinstated saying that hey so I said I'm sorry I forgot to do my activity. These are the things that I did because last year was a whole mess. Um I try to u do my founder job. I try to bring in a youth tech in there. We gave seven weeks of free summer program for high school kids **[11:33]** and I was super duper busy with that. So, I totally forgot, but that was my bad. Um, so this year I'm starting to do something more to the community. Let's see how it goes. So, >> yeah, it's it's it's fun. Actually, I love doing this because it connects me to um where I want, you know, how if if at all, if you ask me 20 years back, if somebody could have helped me, I would have been in a better place today. I could have had a proper mentor for eight **[12:04]** years of my life. I experimented with different roles trying to understand the entire systems you know so but then you know I don't regret it everything was a learning everything was a very very calculated professional move that I did uh to learn a lot of stuff uh and I'm still learning I'm not I'm not an expert in anything I'm still learning we're all learning I guess and that's what keeps uh keeps u you know keeps me awake at night so if there's something new I just jump jump onto it. I'm like, what is **[12:36]** this? How do I make it? And for me, I'm a very, very visual learner at first. So, I need to put my pieces. I need to write up. I still do write in paper and pencil. If I don't understand stuff, I have to sit and think, oh, what is this? How do you do this? I am very visual. I draw stuff. I that's how I understand. And once I understand it, it's very easy for me to communicate with people, too. >> Yeah. There's uh about that that pen and paper. Uh I mean, I I said that one. I do this once in a while too and I I just **[13:06]** felt like I stone hold when I was saying that but like there's research about uh the fact that and and I can't remember exactly but it was about the fact that like when you writing down like with the proper pen uh on paper and I don't think there there's the same effect like with a iPad or something but I don't know I can't remember if it's related to like the feeling of paper the fact that your brain first writing is something it helps your brain remember better or learn better instead of like typing on keyboard and I I wish I would remember **[13:37]** right now why it was the case but uh I mean go on Google search just trust me uh people think there's research about the fact that like when you write on paper but like on real paper uh there's something about it that makes your brain do better at learning and and memorizing and understanding stuff. So it's why like many people outside even outside of text like when you want as you said when you want to learn a concept sometimes it's easier to go on a even on the board even if it's not a real paper but like writing on the board or this is why like **[14:07]** psychologists tell you like hey if you write a journal like like buy a paper one because that helps you probably better process your emotion and stuff like that. So I think it's just a great uh a great way to uh you know try to understand things which is also a benefit of working in tech. I feel like one of the thing I like working in tech uh is that you constantly need to learn because it's going so fast that if you don't continually learn new stuff, you're just going to be left behind uh because going so fast, especially right **[14:37]** now with AI and all that stuff. So I like it. There's so many things I would like to say because you said a lot of things, but uh let me try to to cherrypick some of those things. Uh the first one um would love to talk uh a little bit about like uh actually you said it so maybe you don't have that much more to say but uh Google uh the Google developer expert uh recognition from Google because um I know a couple people that are uh expert on Google but I don't think it was ever a topic of discussion in Deal show and I love those **[15:08]** kind of like expert MVP or recognition program like you you're an expert in specific dominant or you're recognized by the company by being what I would call a virtual evangelist for the company like so not >> working as an employee but still improve the community improve the the whole ecosystem so um h how is it work I know you already said a couple things >> so it it it does it does require somebody to invite you okay there is a application process and even prior to **[15:41]** you applying you should have been involved with the community Um if I think I have a blog about it on my Medium post um I think three years back that I wrote. If anybody is interested just go read it or shoot me a message in LinkedIn. I am very active there. So I will I will tell you hey exactly what you need to do. But you need to do uh they they need to know how much have you contributed to the community. Right? So when I started up even before I applied I think I was invited for a program **[16:12]** called as Google developer expert. um uh road map it was for it was through women techmakers um in 2022. So um I was invited for this program 2022 yes 2022. So I was invited for this program um prior to that I have given you know few talks already. I was around the community. I was also doing events. Events help you a lot because you connect with a lot of people too. Yes. I I never had the intention to become an **[16:43]** expert. However, uh they gave me a ch chance. I'm like, "Oh, you have you know because I love teaching and I I love speaking. Um I gave talks around a lot of communities especially in the Google those back in the days it was called this Google developer um campus groups. I now it is called as GDSE that's that was uh the term that they use. Now it's all Google developer groups on campus. So I have done a lot of onampus stuff either virtual or going there and **[17:14]** speaking as well. So um my because I did all the work prior uh if you go to my website and there is a talks page I think and it gives all the information of what I started. I did a lot of mistake. I was not super duper experty. I was scared to at the beginning but now it's like okay even if I make mistakes you know I just say oh yeah I made a mistake so I move on >> that's fine >> so um initially I was very hesitant **[17:47]** everything had to work perfectly if I even fumble I I used to you know do a lot of judging self judging myself um however it's just that somebody thought oh you're doing very well why don't you apply for this I'm like okay so the GDE road you know the road to GDE program I was also involved and this is uh I think one person among globally was chosen. M >> so yeah uh I was so happy that you know it's like somebody recognized oh I was thinking oh maybe you know somebody is **[18:20]** recognizing I I think I do have some skills right so coming from a non-traditional area so I felt that but then you know what when somebody calls you and says that hey you are you know you are you know you are visible you are doing something to the community and we like you too right >> so uh I was in the program. So they do have an application. Uh if you go to my blog, I I put pictures of all the steps too. There are four or five. I don't know it might have changed. First thing you need **[18:50]** to pass the screening. So the p to pass the screening itself um once it you pass it. Um and there are two more three more uh interviews. one with the community person, one with the product person from from the community itself and then the last one would be with a Googler. So there are four steps to it and the road to GDE program helped me as how do I do how do I make my application stronger? **[19:21]** What how do I do this? Right? I was all over the place. So I came from web from the web space. I love React. I love Angular. you know I I kind of you know because the time that I was working uh there we had both Angular and React so I was like oh no I find this you know and any new thing excited me I'm like okay let me try this and then you know that's how they said no no no you cannot go around this place just pick one particular technology that you like I **[19:51]** said I don't want to be in one particular framework I don't want to be in super duper angular or I and I I was Flutter is something totally you know it's not going to be used in my in my workplace so it'll be a totally different domain for me or a different platform for me. So I said I want to remind in web but I don't want to pick any frameworks per se. That's when um they said okay why don't you go in web technologies there are four options now the categories are totally different now **[20:22]** everything is web AI um so uh I started off with the web APIs the the uh progressive web app that was the category that was in there so I was working on the browser APIs I did a lot of talks around that I did a lot of writing around that please don't you know my previous blogs would have been very very Um it it is all my experimentation as because I first >> and the best part is if you don't understand something you try to learn it **[20:54]** and write a write a blog about it and talk about it. That's that's the whole point of me doing the same rinse and repeat thing if any technology comes up. So even now the the AI thing is a big hype. I'm trying to do the same thing. Okay, what did I do before? How could I do differently this time? Right? So that's one thing that um I did and then the program um there there was a mock application they showed it to us and I had a mentor I I should uh I had couple **[21:26]** of mentors along the way one was uh James um he was he is in um he's in UK he James Milner that's his name he was also he is still a web GDE and he was my mentor and he was guiding me through along the whole process you should do this, you know, that that's something that I have never had in my career at all. >> So, it's not that somebody handheld me. He was just pointing me maybe, you know, this could do better. This could do better. So, I was like, oh, can I do this? Can I do that, too? So, I was all **[21:58]** I I get very hyped up and if you if you see me going, you know, going, >> you become a cartoon, so don't please don't mind my facial expressions. However, I get very excited about this. Um so I do the I did this. So that's how the GDE happened. Let's not talk about that. That's that's totally a different event. Coming to the events, events are something that is that connected me to a lot of people. So it does not happen overnight. Like I said, um the GDG San **[22:31]** Jose LinkedIn page is created um by me. So if you see that there are 500 people, you can also follow me there. But the best way is to go to GDG San Jose um you know Google's page. So yeah that's the page. So if you go there you can join us and you will get all the updates instantly even if the co-hosting ones and you can pick and choose what events works for you. Uh there there is a campaign that's going to be coming this I think starting March. Everybody I know **[23:01]** we are all working on the build with AI thing. So it's going to be coming um with new whatever new stuff being in the community also gives you an access to closer to technology whatever Google is building um it gets you closer but becoming a GDE gets even more closer let's let's put it this way you you have an access to um the information prior of course you are on uh I'm on an NGA I cannot talk about it uh however you can **[23:32]** suggest these things Hey, I tried this thing. Maybe this feature could be, you know, we want more of these feature or we want more of this or maybe I'm having issues and maybe you know all the suggestions that you would give at any you know you do that with any company level too right it's similar things and it's that's something you know you you as a person grow along with tech as well as helping others grow as well. >> Yeah. And know that's one of that's that's one of benefit. It looks a little **[24:04]** bit like the other program I know being part of or help to to manage or create and um I see there's benefits from the company but it always need to be a win-win situation and and this is one of the benefits where like you have access to the product team you can give feedback you have more information uh hopefully there is a lot of so support for you to doing those kind of things like speaking in advance and and starting the user group. So I love that. Um I want to go uh talk a little bit about about those things a little more but just before I want to say hanoj uh **[24:35]** hey Fred hope you're doing well finally here after a long time. Yes that was a long time my friend I was worried about you I was wondering what was happening uh you didn't come for the last couple of episodes but uh happy to see you here always nice to have some friends on the show and uh yeah hopefully uh you're having a great day. So um on that note uh Deepa u again you said so many things that I would like to commend or hat stuff on it. Um first I like the fact that like you know **[25:06]** you're running a meetup. used to do a lot of meetups. time I do damn I I I don't run meetups anymore just because um I do so many of those things part of my day-to-day work that now on my free time I'm more like you know what I don't really want to do meetups or or speak at events in my free time because I do that for I love it but I do that for work and I have other passions but um it's always something that you can do like as an individual if you're looking to you know build some expertise if you want to go within Devril if you want to connect **[25:37]** with the community if you to give back to the community. Uh create a meetup and depending where you live, maybe there's not a lot of opportunity because there's some cities where like there's so many meetups that like there's no space for more. But at many places, uh there's probably like a niche topic where like there's enough people in the community that would be interested about that or a more generic one depending where you live and and creating a meetup. Yes, it's it's work because you need to find venues, you need to find speakers, you need to do promotion, but I think it's rewarding work and I love what you're **[26:08]** doing and I'm going there first to encourage people to create meetups if you have ideas. Uh it's it's it's work. It's a little bit I would say easier than people think. Meaning that people like there's a lot of work, but like once you start doing it, you have hopefully like some other people is going to be able to help you. some local company who may help you to do venues. You don't need to do them in person also. You can do them online which make it a little bit easier in terms of venue or or finding uh those activities too. But what I like one of the things you've **[26:40]** done and I didn't see that often is that kind of collaboration you have with other meetups uh where you just like organize things together. So that helps both organizers because it's one event instead of two. uh that visibility to different topics that that helps at so many levels. So I just wanted to to I guess I don't really have a question. It's more me blabbering for you to like to tell you. It's great idea. It's great. It's I love that you're doing that. >> This this came up only last year, Fred. It was not it was the new thing at **[27:10]** Google. So we were all expecting we were initially creating external links to one chapter. So what you would do is manually go take if I if somebody is hosting an event in GD GDG New York, right? And I like that it's a virtual event. I I personally want my chapter members to know about this topic, right? So I used to create an external ticket and then it'll it'll reroute to their page. It was a big hassle. Then then we were all saying that is there a way that we could just click one button and get it get it done, you know, and now we **[27:43]** don't even have to do it. If you already open to a co-host uh you know category uh so if it's a co-host uh option is anybody enables it it it relays it to every other chapter so it's not one particular um area and having said that if people want to experiment I would strongly suggest try to do uh an online one see how you can build your community first so uh it could be you can join a GDG but if you want to do it, you know, **[28:15]** uh solo or start your own along or along with your friends. Do a uh online first and see how it how it goes because I started during the pandemic. So I had to do only online events to begin with and the first year yes the first uh no I did not get much traction. It was like 100 or 200 people think slowly it grew. So once I hit the mark of you know yes we did a one uh you know uh in-person **[28:45]** event that's when people like in person for sure but then the logistics trying to create an event space and Google does have some you know does have some um uh guidelines or they have their own um I would say there are some logistics to it so now we don't do anything at the Google offices anymore Initially earlier we used to do it now they they I know they want us to do it outside of the campus which is totally fine but still **[29:16]** we have to follow their guidelines making sure that everybody's over 18 and if anybody is hosting in at the office they have to bring their ID so uh in a if you're doing it on the site you don't need an ID as long as they are 18 years old that's that's fine and even if minors are coming they need to make sure that these days kids also are interested however ever um they have to come with their parents. So that's another thing and >> I would nothing happens overnight. You **[29:47]** have to plan it accordingly. But when you have a system in place, I have a document. I just have to go it's now I can do it uh you know for my sleep too. It's not a big it's it just takes some time to do it like how Fred you're doing it with Streamyard and you know uh doing it in Twitter all those things. But one thing is uh I'm not as proficient as red to streaming to different platforms. Uh so we have our own platform at uh Google. So that's how I do it. And you **[30:18]** can also live stream it on YouTube directly there. Um I do have a GDG San Jose YouTube channel too. But it's it's initially I used to download everything and then upload it. Now I don't do that. So it's just a button away. So they made it easier too. So it's this is uh this is something that if you are starting something on your own see whether Google does require certain specific timelines **[30:48]** of doing the events. So you need to be able to contribute to the community at least tw you know once in twice two months. So if you can um you know if you have the time if you can do something you can join the Google developer groups near you or you can go you can start your own you just have to contact and then they will definitely let you know and that's another thing another part of being in the program is also um as a organizer Fred there are when I say **[31:19]** campaigns they we also have free cloud um certification campaigns too. So what it would do is people can come in and you get credits for that. It's not that there might be a couple of raffle for taking the exams for free but that is not but you get all help from uh one of the cloud experts to go through the certification. So that's something uh valuable if anybody wants to just look for you know GDG **[31:51]** um any GDG that is organizing a cloud um certification program it could be online it could be an inperson if it is closer to you go you know go uh take advantage of that and learn the skills so the other thing is any now the build with AI campaign is coming and then they do give you a Google cloud credit for you to try out that particular workshop and that's that's again uh you know something that you can try it on your own that if you **[32:21]** you don't have to sign up for with your credit card or anything because Google does give you some credits for using the tax AI and other stuff. So that's that's another thing. Having said that, um you want to do something on your own, starting on on your own. Um start up with a small meetup group online or your local community and you build it over over the over you know course of time and **[32:52]** >> you'll connect with a lot of people. The reason why I enjoy being at uh the Google developer groups is because I'm closer to u the community and the professionals and also the technology. So there are three parts to it and of course being an expert also helps a lot getting to know the product more and there you know if you want to grow in that area you could also do that. There are different programs a IML they have and Android they have flatter they have **[33:22]** even cloud and different technologies just go the know go check with the experts but you need to be invited for the experts program >> um by somebody and they would ask you I do get a lot of people who who come and ask me hey how do I become an expert I said please read this blog first if you have any questions please come back to me because there's a it's it's it's it you need to do the pre work before you even apply to it. Somebody has to refer you then they should be able to see **[33:53]** whether and please do not do these um AI generated videos on YouTube and say that I've created a channel I have different followers. No, that's not authentic. They want to see your face. They want to they want to see how authentic you are >> and that matters a lot. Even if you are you don't need to be great at anything but try try something. I know if you go to my videos, I'm not proficient either. I I still, but I still want to do it because, you know, I someday I'll get **[34:24]** there like Fred, but I'm still trying. >> So, no. >> Yeah. >> Awesome. Love to have. >> No, I need to I need to learn from you too, right? So, I I like to see how people are doing and we all learn, right? >> We all need to learn from other people. So, no, there there's a lot of bit a lot of benefits to those programs. It's a great recognition, but there's benefits to and it's fine if you're not part of one too. I mean like because they're there limited spot. Um I guess also my my question my point was like yeah if you can build something if you can create stuff fine but also if you don't **[34:56]** have the time you don't have the energy you're not sure where to start uh just go to those things. to go to meet tops, go to the events, you meet people, you create connection, you don't know the person you met at an event, if there will be potential uh someone who's going to hire in their team in the future in your future bus, future colleagues or whoever. And it's just just a great opportunity. So there is more to those meetups and events than like the the technical content by the speaker or whatever the content the speaker is sharing. So with that note, um we could talk forever. Uh I have so many other **[35:27]** things I would like to discuss with you but uh I think it's time for us to move to uh the second part of the show which is the uh demo and um I have a huge memory blank right now. Uh I forget like I I know it was something about AI but I'm so sorry it's totally unprofessional but like I really don't remember what was the deal. >> No worries. So it was multi- aent. I I saw that there was a lot of things with uh you know lang chain and a different backend base. So I thought okay maybe **[35:59]** let's let's do something different um and do if if you have to be a you know uh been if you have to be in in a full stack environment and you want to code a code review pipeline how would you do that? It's just mimicking um in a real life scenario. So I'm going to be let's see how I can share here. Uh I'm gonna Okay, I'm not going to be seeing you, but uh I'm going to minimize this thing. >> That's fine. You're not missing so that >> Okay. So I'm going to sorry. So this is **[36:32]** what we going to be doing, right? So if you see here in a real life scenario, you will be just again know the developer goes pushes and then there is a VR open CI is triggered, right? And then what this demo is going to be teaching know I'm going to be talking about is how do you add a AI agent to the pipeline. So you can add three agents. I chose chat you you could use uh you know cloud for everything but I chose to show two three I know two different um um um AI providers here. uh **[37:06]** one is uh you know the reader agent from open AAI because it's good for summarizing and uh summarizing and I'm using two agents a reviewer agent and the fixer agent. So in a real life scenario the agent would go check for your code. I'm going to show this in a real quick moment. Before I do this let me uh I have to see what I did this. Okay. So, where is my thing? So, **[37:37]** bear forward. Can you see my code? >> I can see your code. >> Okay. Cool. Uh, so let's start with the rout. It's I'm not going to be going too much into it. You can everybody can go do this. It's very super easy. But I wanted to show that we can do it for backend and front end scenarios. So, this is the route agent where I have it for backend and as well as for front end. And there is a reader agent where I'm using OpenAI uh GPT model for uh O mini and then it **[38:09]** reads so what it it it can summarize it. It just summarizes what the code has given right and then the second one is uh the reviewer which is from cloud it's using cloud sonnet and it just uh reviews the entire code and sees what are the potential issues here. Third uh one is a reviewer one which is also from oh sorry the fixer one which is also from clot and here it just it does not fix your code but it does say that hey there are some issues. This still would **[38:41]** need a human eye to see whether you can push it to the pipeline or whether you can u you know merge it or not. So this is some and I created couple of scenarios I think seven of them I think. So uh let's see the backend scenario um see these are the agents that I'm using uh reader agent and then the reviewer agent and the fixer agent and the backend scenarios how do you do if you do a hardcore this I have seen it very commonly so if you're running tests or **[39:13]** something people do run this uh and your information is getting you know exposed so in that case howard uh claude claude and uh GPD catches it right so that's that's another thing and API key is exposed. We I do this very very often when I'm doing the demo one. I usually when I try to use uh uh you know when I'm using uh you know I forget so I forget to create it uh when I'm testing **[39:45]** it I just directly enter the API key and then I expose it to so then I realize oh my god I forgot to create a environmental file right so this happens quite often and this is um this would also the backend one also tries to create a mimicking uh SQL injection vulnerability and then the login one and also a ne for performance um based if it's a nested loop it's going to show that hey there are three uh this is **[40:16]** going to be uh work n3 so this is a performance disaster so this is going to show and then on the front end side um it is going to be um selecting a card and then how is the how is it uh missing the there is a missing um um alt text and then there is this I have seen you know the the layout changes for example if you're using RTL languages um to the Latin one so how does it fix it so this **[40:48]** is what there are some scenarios and the color and stuff like that so I'm going to be running this demo real quick if you have quick time in here and then you can take a look at this also and let's see how we can run this so So I'm going to run the first the reader agent for the back end. So for example I'm running it hopefully it is running. So it does say okay it does summarizes it saying that hey this is the typescript one. So it is connecting to the database and then we are connecting to the middle **[41:19]** layer and then the API endpoints are so this is done with uh chat GPT uh the GPT uh for uh model and then the reviewer agent it looks for okay there are some issues. So there is a SQL um thing critical issues and it also says that there is a hard-coded secret and once it's all done it just it gives you the fixer one does not fix anything but then it says that you could you could look through it. This still requires a human **[41:51]** eyes. So make sure that whether these issues that are uh reviewed by the reviewer agent uh somebody could go and fix it. So it would fix it but still needs somebody's eyes in there >> right. So this is one part. This is done. And then what I found okay this is not relatable. It had hallucination. Then uh no I I tried to give it specific uh examples here. Uh it was hallucinating. Then I had to it is not right the it was not right the first **[42:22]** time when it uh the reviewer agent had issues. it was hallucinating when I was using anthropic the first couple of times and I said hey how is this you know you're saying that it's a SQL injection vulnerability but then it's just testing it you know using for it it it had a issue in here and let's go to the front end one so what it it give you um I I made it in such a way that it can show the visual part of it too so again it tries to create look for the **[42:56]** um product card with the with details and if you see there are some I personally asked for if it was using a mobile how the layout changes and what are the what are some of the things it does say that there are there are no critical issues but there are some issues here just give me one second while it processes it and then it'll show exactly and uh What one thing I found out was if you have done some um **[43:29]** uh internationalization there are some ways that the numbers are displayed in different ways. Hang on. Let me see. It was working last night. Trust me, it was let's give it a couple of go. So yeah, this so if you see here um it checks for the vet and and the pricing, right? This is in English. Uh it is it is okay. It is it is like this the beginning and then it got fixed. So **[44:00]** again the layout itself has uh has been fixed too. So if you see Arabic the initial one you see the the number the currency has not changed here and depending on the browser language setting because I was when I was testing this whether this you know the numbers were not like this right it was giving something in Arabic so I don't know Arabic but then I saw that hey why is this numbers or not numbers I could see a number and a text difference right so **[44:32]** depending on it did not Even cloud did not catch it. So I said, "Hey, you know, this is something wrong." So don't trust with what the agent says. You have to know whether it's doing the right thing or not. That's why I said that you still need a the fixer agent would not fix anything. You still have a you need a human intervention here. So like I said, then the text have been changed and also oh the the German one here, Germans write the currency in a different way. **[45:03]** So the N is at the the euros at the end and if you see there's a comma in here and little and everything is also um in German here earlier it was not. So these are you know the common things in uh front end um development you would use and then the back end backend one of course you can try to play with it you can this is an agnostic one I just chose um uh no open AI and anthropic however you can use it you can use u anthropic **[45:36]** for for all three however um open AI's um um model is cheaper and it it is faster faster for in terms of summarization. So that's why I want to just choose this. If you are working at a place where you know if you're working at a company and they can afford then you can use all three as cloud or you can use all three as open too. So that would be my demo. >> Yeah, >> hang on. So that was quick. If you want to take a look at it, it should be in my **[46:06]** in my GitHub >> wonderful. If it is not there, I will push it because I forgot to I will make sure that my API keys are not exposed. You this is a paid version. So, uh it's it's not that much. You just have to pay $5 and so that you can experiment it. I've been experimenting. >> Yeah. Yeah. Still still we don't want your uh your your API key exposed. No, I like that because uh there's uh things that happen with AI and and we're going to close on that. Uh but you know like there's different cost. It cost a lot of money if you if you use it a lot. So **[46:37]** like trying to optimize for different models makes sense. So uh it's like I I feel hope like it's like when we're talking about cloud you know like when cloud would becoming a thing and we're like no don't just put all your eggs on the same cloud like you use multi cloud approach because they have like like it's a little less true right now but like before each had their own pros and cons and like you could use like a specific cloud for specific task and another one for another task. And anyway, it was just a great way to be sure you don't have like a single uh what do you call that like a single **[47:08]** entry failing entry point something that so I guess that's the same thing too here and it's optimizing like the cost and and using the best model for what the task you have to do >> you could also even prior to using this model you can also use something because there are some static ones called a sonar cube or something uh like that you can also uh use that along with the pipeline so that you don't have to you know use so much of your tokens too. >> Yeah. Yeah. And and I love also that you said that I have funny enough I'm **[47:39]** speaking on the meet tops tonight about AI and uh one of my big obvious takeaway of my talk is the like there's some benefits of AI but there's also some risk. There's also things that that people don't talk about enough that we should talk about. But uh one of the big thing that I want people to come back is like like never remove the human from the process especially on right hand. Maybe in the future we'll see but right now Alice Nation uh sometimes the LLM are trained on old data. Uh you cannot **[48:10]** like they're doing a pretty good job but you cannot like bly trust any agents right now. So always have a human being part of the process to be sure that what's fixed, what proposition, what's change makes sense. Uh I think I think it's really important. So it may not seems like you're going to be as productive is like if you automate everything, but at the end of the day, I think you're going to end up with better quality and it's going to be better in the medium long term. So >> and also like I said, this could be agnostic, right? So you can you can swap it with llama or you can swap it with Gemini. But then I did try for with **[48:41]** Gemini too. I I I ping you. I told you to. So if you uh I was trying the Gemini 1 no 2 and 1.52 I was trying testing it was not working the reading reader agent was like okay maybe I and you need to use vortex AI for that I tried with one of my free cred and the credentials that I had. So if you have the time and I I would say that if you're familiar with GCP and Vertex AI then go with it. I wanted to show a demo that if somebody **[49:12]** wants to show it, if somebody wanted to use it, a very very basic uh plugand play, this is what this would be easier. >> That's awesome. So, uh thanks for the demo. Thanks for uh sharing your your passion. Uh I think people saw it. You're really passionate. You're really excited about things which I love. Uh that also makes my life easier as a MC because I was like I was asking you a question and you were going and like like like talking about it with passion and everything and it was that's amazing. Uh I'm not even joking. That's great. I'm not even teasing you. I love **[49:43]** it. Uh the teasing part is that there's rarely people that talk more than me and you're one of those persons. So uh you decide if it's a good or a bad thing. I like it because I talked a lot. But uh I mean that was that again uh thanks for being in the show. That was amazing. Is there any last thing you want to leave people with before we uh close that episode? >> Um well I just started a newsletter. Uh so this is also one way that you could start with the community if you want to. I started with a newsletter and if you I **[50:15]** I am I'm going to be focusing more on AI platform and architecture on my newsletters. Um so if you want to sign up it's there on my LinkedIn. I just started this uh so yeah and I would also be running few uh from March 5th I'm planning to run a forum just come in it's going to be free it's going to be more of AI architecture platform all those stuff so I'm you know it's not it's not hands-on work I'm not going to do but then we can discuss like how we talked about today it's bug we we talked **[50:46]** about this we could also talk about how what are the things that breaks or how is the cost going to implement you know cost going uh uh how it's cost effective or what whether should I use it whether should I not use it all those stuff and how do we architect and what are the mistakes that we go along the way we can talk all about it too >> that sounds good so you're a busy woman so uh thanks again for being on the show was a pleasure having you uh thanks everyone who joined us live thanks to the people also who's going to watch the **[51:17]** recording and uh I'm not sure if we're going to see each other in two weeks because uh I think In two weeks I'm going to be in Europe. I'm going to be in Amsterdam for Laracon EU. Also speaking at a ride the docks meetups there and you know getting busy. So I won't have an episode in two weeks but I think and I do not know my schedule but I think I'm going to be back in about a month with another episode. So thanks everyone. Thanks Diva. Have a good rest of the day and see you soon folks.