Gaoang Wang

When preparation meets process gaps 🤔 I spent a week preparing for an interview: system design, algorithms, and deep dives into my past projects. Then the first question was about sponsorship. I shared transparently: I am authorized to work in the US for up to 3 years on OPT, and I would need H1B sponsorship in the future. I had also disclosed this during the application. The interview ended immediately. Disappointing, yes. But it reinforced something important Clear alignment upfront saves time and keeps hiring focused on engineering ability and impact. Here is what I bring to engineering teams Scale and reliability 🚀 Scaled distributed systems to 50K plus daily requests at 99.9% uptime ⚡ Improved processing speed by 92% with multi threaded Java pipelines 💰 Delivered 1.2M in annual savings through performance optimization Cloud and modern engineering 🔧 Built production systems using Java, Kotlin, Python, Node.js, Spring Boot, AWS, Kubernetes, Docker, Redis, PostgreSQL 🛠 Strengthened delivery with CI CD, testing, observability, and security hardening, including eliminating 500-plus critical vulnerabilities AI and applied learning 🤖 Built an AI Q and A chatbot using Llama 3.1 on edge serverless infrastructure 🎓 MS Computer Science, AI specialization, 4.0 GPA Leadership 👨‍🏫 Mentoring 200 plus students in Database Systems and OOP, running labs and code reviews Recent builds • Distributed URL shortener with under 50ms p95 latency at 1K requests per second • Crash-safe key-value database in C with 20K plus ops per second • AI powered Q and A chatbot with persistent conversation history I am open to Software Engineer roles and can work on OPT for up to 3 years. For the right long-term fit, I will require sponsorship in the future. For recruiting teams: if sponsorship is not available, confirming it before interviews helps both sides invest time meaningfully. #OpenToWork #SoftwareEngineering #DistributedSystems #CloudNative #AWS #Kubernetes #ArtificialIntelligence #OPT

36

1 Comment

** Big update on the roadmap for Meta's MTIA roadmap (our Meta-designed AI chips) ** We just shared today a big update on the MTIA (Meta Training & Inference Accelerator) roadmap. The main takeaway is we're accelerating the whole set of programs to get rapid iterations on chips (given how fast the model needs are evolving). 👉 Overall blog is here: https://lnkd.in/gaHuxUfg For the networking aficionados, you can see a whole lot of networking that we've been working on inside MTIA - we're excited to finally be able talk about it! 👉 The scale-up and scale-out networks we've already talked about. It's sort of buried in there, but we are already running distributed training clusters with MTIA 300 for core ranking & recommendation (even as we focus future generations on Inference). 👉 The communication library (HCCL) that is purpose-built to take advantage of networking-specific capabilities we've built into the chips to offload the main compute. 👉 The in-built support for fused compute/communication kernels to minimize latency (key for both training and inference workloads). Super-proud of the whole MTIA team!

143

4 Comments

A friend of my colleague from Google, who’s based in the US, has just interviewed for the E4 role at Meta. The total package on the table was up to ($260-270k ~ 2.7 Cr) Sharing his interview experience for everyone preparing for interviews right now: 1. Technical screen – pure DSA - 45 minutes - 2 problems straight from Meta’s “Top 25 in last 3 months” on LeetCode - Code could run, but there were almost no test cases - He first explained the approach out loud, wrote it as comments, then coded and walked through examples He cleared this, then got the final loop. 2. Final loop DSA round - Another 45-minute DSA interview - Difficulty very similar to the screen - This time code could not be executed, so clarity of reasoning mattered more than speed - He still did 2 problems, 10 to 15 minutes each, and spent the rest of the time explaining trade offs 3. AI coding round - Given a small codebase with functions to fix or implement - 4 tasks, he fully solved 3 and explained the approach for the last one - AI could help read the code and give context, but not solve details, so fundamentals still mattered 4. System design - 45 minutes - One of the most common Meta questions from HelloInterview - They used Excalidraw, he followed a strict delivery framework and kept asking for feedback as he went - Clear trade offs, scaling discussion, and follow ups on high traffic cases Before we move forward, I also want to tell you about our latest addition to Layrs. The AI based interview tools for system design and DSA are live and free to try: layrs.me/interviews You can run DSA mocks that pick real interview-style questions, get follow-up questions step by step, and get instant feedback on structure and coding. For system design, you can practice delivery on real prompts, draw designs, and get nudges on coverage gaps before you sit in front of an interviewer. 5. Behavioral and the surprise twist - 45 minute behavioral - He under prepared, thought “tech is what really matters” - Feedback after debrief: very strong technical signals, but too many missing signals on behavioral - Meta asked for a behavioral follow up - He then spent a week building strong STAR plus learning stories, did a focused follow up round, and only then cleared into team matching The point is simple. For roles at this level, you cannot treat DSA, system design, AI coding and behavioral as optional sliders. All four show up, and any one of them can decide the outcome.

1,087

18 Comments

Meta Quietly Delayed the Release of Its Flagship AI Model Codename “Behemoth.” ↳ Not once. ↳ Not twice. ↳ But possibly 'til fall or later 🤷‍♂️ Why? Because it’s not ready—and may not be anytime soon. WSJ 🔗 https://lnkd.in/eARKVQdU TL;DR • Meta's leadership is concerned about the Llama 4 team's performance. • Engineers are struggling to enhance Behemoth’s AI capabilities to meet user expectations. • Despite $72B+ in annual capex, the model's hitting the ceiling—fast. Early in its development, Behemoth was slated for April release to coincide with Meta’s inaugural AI developer conference. Meta put out two smaller models ahead of the event, but later pushed the target for Behemoth’s release to June. Now it’s been delayed to fall or later. The two models that were released in April initially performed well on the popular AI leaderboard, Chatbot Arena. However, it was later revealed that the model submitted to the leaderboard wasn’t the same model that was released to the public. Meta previously drew praise for its velocity in catching up to rivals in the global AI arms race and setting the OSS precedent (even though it has guidelines for commercial use). 🚨 This isn’t just a problem with Meta—we're seeing a wider industry trend known as "the LLM ceiling." Status on GPT-5? Delayed. Claude 3.5 Opus? Still MIA. Across providers, hallucinations remain unsolved. Companies are pivoting to focus on agentic workflows. Some researchers see the pattern as evidence that future advances in AI models could come at a far slower pace than in the past, and at an astronomical cost. For the sake of progress, let's hope that's not true... (as I log into Gemini) _______________________ Want more AI by Design? Blog 🔗 in comments. ♻️ Reshare if you thought this was helpful. 🚀 Join a growing community of PMs who launch and scale 0→1 products with AI, leading with clarity and confidence. Subscribe 👍 link in bio.

9

1 Comment

4 years at Uber today. My most valuable lesson: grow people’s Data Science and TPM skills to supercharge collabs. This is how I did it 👇 HOW: - Composed a 100+ page document called “Data Science For Everyone”, which is equipped with step-by-step tutorials, videos and notes that accelerate and orientate practical skill development for real problems - Maintain the Guide current with regular updates, engaging visualizations (where possible) and notes that complement Source of Truth documentation elsewhere -Circulated passively and engaged comments/suggestions inline and via slack, encouraging others to share if they found it valuable - Made myself available to chat, zoom and present IMPACT: - 1.2k+ users across the business - 50+ teams, involving professionals such as: MLE, SWE, Data Scientists, Data Engineers and TPMs/PMs across Safety, Risk, Legal, Strategy & Ops and many others - 400+ distinct users per month for the last few months PURPOSE: - Grass-root tool to onboard new hires - Teaches how to use GenAI tools - Architectural designs for experimenting with simple LLM use to multi-agentic RAG workflows - Generic code bases for people to get started - Standard project directory designs that illustrates best practices for scalable and efficient growth - Guidance and code snippets for how to use Python for big data projects, including how to build automation and classic ML solutions CIRCULATION: - Organic - Met a need shared across professionals across all levels. NEXT STEPS: - I hope to collaborate with others to compose a Data Science course curriculum for K-12 students - In the meantime, I’ll be working on general educational resources A high tide rises all boats and builds a meaningful network of collaborative ships!

80

10 Comments

We're hiring three AI Analytics Engineers on the Data Science & Analytics team at Airtable. These are early to mid-career roles for builders who are excited about working at the intersection of data science, analytics, and AI. One of these roles will report directly to me as we spin up a new AI & Analytics Platform function. Things are moving fast. The most impactful work in data right now isn't building models — it's fundamentally changing how a company gets answers from its data. That's where my focus is, and it's why Airtable is investing in a new team to own it. Over the past six months, our team stopped experimenting with AI on the side and restructured around it: - Agent-powered development workflows — I built shared AI systems for the data team embedded in how we write SQL, build pipelines, document business logic, and debug. Team members are regularly seeing 2x to 10x speed improvements — not from shortcuts, but from AI woven into every step of how we build. - Conversational analytics — Ad-hoc questions that used to fill our queue from leadership, product, GTM, and finance are increasingly answered by an AI agent — in natural language, where people already work. -A context layer that makes it trustworthy — We built a system that gives the agent real business knowledge — which table is correct for a given analysis, how metrics differ across teams, why something spiked last quarter. Context that usually lives in an analyst's head. - A learning loop that compounds — Every user correction becomes a permanent improvement, not a one-off fix. — We're doubling down. Airtable is standing up an AI & Analytics Platform function within DS&A to keep building at the intersection of data science, ML, analytics, and AI — and I'm leading it. We need more builders. AI Analytics Engineer, AI & Analytics Platform First hire on this team. You'll build the context layer that makes AI tools accurate, design evaluation frameworks that make them trustworthy, and develop agent systems that let the org self-serve with confidence. You'll work directly with me to shape what this function becomes. https://lnkd.in/gFstQqQb We're also hiring two roles embedded in other parts of the business: AI Analytics Engineer, Marketing Analytics Own data infrastructure, dashboards, and AI-native tooling for Marketing. Build the canonical data sources leadership relies on. https://lnkd.in/gYHS55ma AI Analytics Engineer, Business Analytics Own financial data models powering ACV, ARR, billings, and revenue. Partner with a Finance team that's building their own AI agents and workflows. You'll build the data foundation they run on. https://lnkd.in/gV5hnJjG — I'm looking for people who build before they're asked. Who have a side project that started as "I wonder if I could..." and turned into something real. Who pick up a new tool and immediately try to break it. If that sounds like you — reach out or apply. I'd love to talk.

79

8 Comments

About a year ago, I moved from leading a Science team at Amazon to a Product Manager role in Pinterest Ads. Colleagues, friends, and recruiters all ask me the same question: “Why the transition?” My answer: the choice wasn’t between Science (manager) and Product (IC). It was to optimize for learning the skills and experiences I believe will be most relevant and impactful in the coming years. I won’t add another “AI is changing everything” LinkedIn post. It’s obvious AI is already accelerating the customer signal → roadmap → product cycle. The boundaries between Engineering, Science, and Product are blurring: Product will become more full-stack and ship code directly, and Engineering/Science will be more directly customer-facing—hearing feedback and use cases firsthand. In that world, a “dangerous person” means having both the technical depth to dive deep into a problem, and the functional breadth for end-to-end execution to deliver results for customers. One could argue this has always been the case, but as AI lowers entry barriers across domains and complements one’s existing expertise, the technical depth + functional breadth combination increasingly feels like table stakes. So I evaluated what I already know—and what I need to learn. What I know: I was fortunate to rotate across different business domains at Amazon and learn how economics/ML/data science are applied to ads measurement, public-policy for AI infrastructure, measuring and improving customer experience (e.g., subscription and omnichannel). Having built and managed a science team, I also became familiar with different operating models for the science function: embedded within product teams or as centralized teams. What I want to learn: 1. Owning customer outcomes: shifting from “Is this statistically significant?” to “What should we do to improve customer experience and drive revenue?” 2. Go-to-market through cross-functional partnership: working closely with Product Marketing, Sales, Legal, Ops, and others to drive strategy and execution. Instead of optimizing to become a “better manager,” “better scientist,” or “better PM” in isolation, I’m optimizing for building my own combination of skills—and finding the right opportunity fit to solve meaningful problems and deliver results. It’s been a rewarding year at Pinterest: more often than not, I’m in the driver’s seat for outcomes for customers and stakeholders, rather than just providing “data-driven recommendations.” I’ve learned a lot about the full cycle from ideas (inputs) to customer and business outcomes (outputs), and I am excited about what’s more to come on this journey!

120

2 Comments

Big news from Meta! 🎉 We just announced a long-term partnership with @AMD to scale AI infrastructure with Instinct GPUs. This is one of the largest AI compute buildouts in the industry. Read more ➡️ https://bit.ly/4qS6U9R "We're excited to form a long-term partnership with AMD to deploy efficient inference compute and deliver personal superintelligence. This is an important step for Meta as we diversify our compute." — Mark Zuckerberg, Founder and CEO of Meta.

4