Bill Chen

✨ Feeling refreshed and energized after a vacation, and I'm excited to share 𝘁𝘄𝗼 𝗶𝗻-𝗱𝗲𝗽𝘁𝗵 𝘁𝗲𝗰𝗵𝗻𝗶𝗰𝗮𝗹 𝘀𝗲𝘀𝘀𝗶𝗼𝗻𝘀 I had the pleasure of recording. If you're building with LLMs and AI agents, these deep-dives are for you. 🤖 𝟭) 𝗕𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝗔𝗗𝗞 & 𝗩𝗲𝗰𝘁𝗼𝗿 𝗦𝗲𝗮𝗿𝗰𝗵 with Kaz Sato 👉 Key Takeaways: How an Agent Development Kit (ADK) works with Vector Search to create sophisticated, production-ready AI systems. 👉 Beyond Basic RAG: We move past simple Q&A to discuss architectural patterns for building powerful semantic search and Retrieval-Augmented Generation (RAG) pipelines. 👉 Practical Implementation: A look at the code and components needed to bring these advanced search agents to life. 🔗 𝗪𝗮𝘁𝗰𝗵 𝗵𝗲𝗿𝗲: https://lnkd.in/gjKvq-MM 📈 𝟮) 𝗠𝗮𝘀𝘁𝗲𝗿𝗶𝗻𝗴 𝗔𝗴𝗲𝗻𝘁𝗢𝗽𝘀 with Dr. Sokratis Kartakis 👉 Key Takeaways: The "Day Two" Problem: Why traditional DevOps and observability tools fall short for monitoring complex, non-deterministic AI agents. 👉 Metrics That Matter: Learn the key metrics for tracking agent performance, cost, and reliability to ensure your agents are effective and efficient. 👉 A Framework for Reliability: Dr. Kartakis shares a practical framework for debugging, evaluating, and continuously improving your agents post-deployment. 🔗 𝗪𝗮𝘁𝗰𝗵 𝗵𝗲𝗿𝗲: https://lnkd.in/gpdC_Jk9 Which topic is more critical for you right now: building new agent capabilities or managing them in production? Let me know in the comments! #AI #GenerativeAI #LLM #VectorSearch #RAG #AgentOps #MLOps

49

3 Comments

Jenny Wang’s startup Alta just raised an $11 million seed round to bring a personal AI stylist experience—think the iconic “Clueless” closet—to life. What’s remarkable isn’t just the funding but how Alta is transforming how we interact with our wardrobes. Users can upload photos of their clothes, forward receipts, or search Alta’s extensive database to build a digital closet. The AI then offers personalized outfit recommendations tailored to lifestyle, budget, weather, and events—down to suggesting looks for a conference or a night out. Alta’s AI doesn’t just recommend; it lets users mix and match new items with what they already own, visualizing choices through virtual avatars. Backed by industry stars like Michelle Obama’s stylist Meredith Koop and investors ranging from Menlo Ventures to fashion insiders and tech CEOs, Alta blends deep technical rigor with authentic fashion expertise. This convergence of AI, data, and personal style signals a broader shift in consumer tech: intimate, context-aware experiences that genuinely augment our daily lives. It also raises interesting questions about how much AI can or should personalize in areas as subjective as fashion. How are you thinking about balancing AI’s promise of hyper-personalization with preserving the human touch in your own projects? #ArtificialIntelligence #ConsumerTech #FashionTech #StartupInnovation

3

2 Comments

Some fun news - I’m proud to announce Northslope has raised $22m to accelerate the end of one-size-fits-all software. Northslope Technologies is the first Palantir-native applied AI company. We ship mission-specific applications that amplify our customers' differentiation rather than conforming them to the status quo. Our products are helping doctors catch cancer earlier, manufacturers build and fly more performant rockets, and energy producers supply more renewable megawatts onto the global grid. We ship our AI application suites on Palantir Technologies - the world’s leading (and only) enterprise AI operating system. We are proud to have been partners since our founding and to be the only Elite partner in Palantir's Vanguard network. Fundraising itself is not an outcome. The outcomes are what our customers deliver on their Northslope applications for their patients, customers, and citizens every day. But this raise is validation that our mission-specific approach to software is working in production and enables us to scale to meet the overwhelming demand we are seeing from customers across industries, around the world. I am so grateful to our customers, who have invited us to help them change the world for the better, and to our investors Colin Anderson & John Fogelsong at Friends & Family Capital, and Adam Ross and Daniel Friedland at Goldcrest Capital for being in our corner from day 1, along with the teams at Fifth Down Capital, Leblon Capital, and the community of Palantir alum angels who have backed us from the beginning. Most of all, I am so proud of what the Northslope team has built in just 18 months, and I'm genuinely lucky to get to work with such an inspiring group of people every day. Come join us. Read more courtesy of Axios and our Press Release. Links to both in comments.

1,150

139 Comments

🚀 LangChain’s leap to unicorn status? According to TechCrunch, AI infrastructure startup LangChain is raising a new round at an impressive ~$1 billion valuation, led by IVP.  What’s driving this surge? ° 111K GitHub stars and robust open‑source traction ° Innovative tools like LangSmith fueling $12–16M ARR A strong footing in the fast‑evolving LLM ops space 💡 If you’re building or investing in AI-powered applications LangChain’s trajectory is one to watch. #AI #Startups #UnicornAlert https://lnkd.in/ddcnHXRd

3

We all hate opening a full fridge and feeling like there's "nothing to eat," so I tried to fix that with a little Python. 🐍🥘 Meet Patatri's Kitchen—my personal AI assistant that tracks my pantry and plans meals based on what I actually have in stock. ✨ I used this project as a playground to dive deep into Agentic AI and Google’s Agent Development Kit (ADK). 🧠🤖 Instead of a simple chatbot, I built an intelligent "Kitchen Orchestrator" powered by Gemini 2.5 Flash. It autonomously handles the logic to: ✅ Analyze live inventory 📊 ✅ Prioritize existing ingredients to reduce waste 📉 ✅ Auto-generate shopping lists for missing items 🛒 ✅ Auto-purchase required items for cooking 🛍️ Just a fun build to experiment with orchestrating AI agents! 🛠️🚀 Check out the demo. 👇 #Python #GenerativeAI #GoogleGemini #AgenticAI #MachineLearning #GoogleADK #Coding #SideProject

22

2 Comments

We are hosting a mixer this Wednesday to learn from our peers in New York how they are thinking about AI agents their strengths and limitations. We will be present sharing lessons from building simulation environments for agents in reinforcement learning workflows. Cya there! (link in comments)

6

1 Comment

🚀 How we built Advance Orders at Groww Thrilled to share how we built a key foundational capability at Groww using Temporal — one that now powers highly deterministic, resilient, and reliable order experiences for millions of users. What began with doubts about using workflow orchestration for trading has evolved into a clean, fault-tolerant architecture that simplifies and enables several critical order workflows — including OCO, TSL, GTT, and forms the foundation for every future advance order we plan to build. ✨ Durable orders that never fail ✨ Orders that auto-heal when glitches occur ✨ Reusable SDKs that let us plug-and-play any new advanced order type From scattered state machines and complex choreography to a unified, workflow-driven system, this has been a major leap in how we design and scale our order infrastructure. 📖 Full write-up: https://lnkd.in/gGuyc683 #Groww #Engineering #Temporal Temporal Technologies #WorkflowOrchestration #DistributedSystems #TradingSystems #TechArchitecture #Scalability #Reliability #Fintech #AdvanceOrders

67

3 Comments

Thinking of joining a startup as a founding engineer in NYC or SF? This is a must-read. “Thriving as a Founding Engineer” from The Pragmatic Engineer features real-world lessons—from being the technical architect to driving PMF. It covers the balance between fast iteration and scalable design, and why a pragmatic, user-driven mindset is key https://lnkd.in/gBbPXh_X . If you’re a senior software engineer eager to own product and tech in a high-velocity startup, this is the kind of thinking that sets you apart. Let’s connect — I’m working with founders who want engineers ready to build and lead from Day 1. #foundingengineer #startuplife #softwareengineering #startups #NYCtech #SFtech #productmarketfit #buildthefuture https://lnkd.in/gn2gPgNQ

4

I recently wrote a short piece breaking down how DoorDash uses Large Language Models (LLMs) to bridge behavioral silos across multiple business verticals — restaurants, groceries, and retail. What stood out to me is that LLMs aren’t being used as a replacement for recommender systems, but as a semantic layer that helps translate user intent across domains where data is sparse or fragmented. The article explores: Why traditional recommenders struggle in multi-vertical systems How modeling intent works better than modeling actions Why hybrid architectures (LLMs + classical ML) are more practical in production How this idea generalizes beyond food delivery If you’re interested in recommender systems, applied ML, or pragmatic uses of LLMs in production, you might find it useful. https://lnkd.in/gTzuB2Bb Would love to hear thoughts from others working on recommendation systems or multi-domain ML problems.

7

Two years ago, I was designing buildings. Today, I’m helping build AI agents. At first glance, architecture and AI seem worlds apart. But the more I work in this space, the more I realize they share something fundamental. Both are about designing complex systems. Architecture designs spatial systems — how people move, how structures interact, how invisible infrastructure supports everything we see. Companies work the same way. Behind every business are operational systems: supply chains, coordination, approvals, logistics. At BackOps, we’re building AI agents to help automate those invisible workflows that keep companies running. Proud to see the team reach this Series A milestone. Check out the announcement video below 👇

13

4 Comments

My team and I partied during McGill’s largest hackathon… and still won. It’s more complicated than you think. At McHacks, my team and I built LookBack: a networking tool that weaponizes context in real life. We even filmed the demo at a party happening that same night… and yes, we stayed. LookBack uses computer vision and AI to recognize people in front of you and instantly generate conversation leverage: who they are, what they’re building, and how you might be connected, all without pulling out your phone or breaking social flow. Let’s be honest, this technology can be uncomfortable. It blurs lines. It can be altered for surveillance, and it is considered a shortcut. But that’s exactly why it’s interesting. The intent behind LookBack is simple, great conversations don’t start from nothing. They start from context, which already exists, scattered across platforms and forgotten under pressure. LookBack brings it back, in the moment it actually matters. The video demo shows LookBack operating on AR glasses as a proof of concept. Because AR hardware wasn’t available, we built and tested the full system using a phone’s camera. The interface changed, not the vision. Big shoutout to my teammates Vlad Matei Rusu and Jacob Leader for an unforgettable night. So LookBack, and never walk in blind. https://lnkd.in/e5kKvauR

210

23 Comments

I pulled three all-nighters building an AI rules linter ahead of our $4,200,000 Seed launch at Weave (YC W25). Here's why: AI coding tools like Cursor and Windsurf work by taking in massive amounts of context at once. If that context isn't high quality, your results reflect that. The best way to optimize this is through rules files that get parsed before every request and fed to the LLM. But most teams have terrible rules files that actually hurt their AI coding workflow. The AI Rules Linter automatically analyzes your GitHub repository and pulls in all your rules files to check them against industry-standard heuristics, suggesting changes for optimal performance. Now, I'm building automatic remediation that will split your rules into optimal file lengths and generate new rules based on your codebase patterns. Think automatic error handling, logging styles, and testing rules that match how your team actually codes. In 2025, rules files are the backbone of our coding workflow, and if they're not good, everything else suffers. Try it out at: https://workweave.dev/

57

10 Comments

In our latest workshop, Alex J. Hong, from Syncly (YC W23), shared how brands can turn fragmented customer feedback into proactive, data-driven support. We covered: • Why visibility breaks as teams scale and signals scatter across tools • How Syncly’s CODA framework, Centralize, Organize, Discover, Act, creates a single source of truth • How AI transforms qualitative feedback into structured, actionable insights Thank you, Alex, for the insights! Full recap in the comments.

30

3 Comments

YC's most common advice to founders: Launch quickly: "Don't wait for perfect. Launch and iterate." Do things that don't scale: Airbnb photographed apartments themselves. Stripe installed their product in person. Find the 90/10 solution: - 90% of the value - 10% of the effort - Ship fast, learn faster Talk to users: "Write code and talk to users. Everything else is a distraction." Don't scale too early: - Product-market fit first - Scale second - Most startups die from scaling before PMF The best founders are flexible but persistent. Pivot the plan, not the vision.

5

1 Comment

Bellevue is becoming the new AI battleground. ⚔️ Elon Musk’s xAI just leased a downtown Bellevue space, footsteps from OpenAI’s new office. Both are setting up shop in Seattle’s Eastside, turning it into a hub for AI model development far beyond just operations. What’s notable? The roles xAI is hiring for GPU kernel dev, world modeling, generative AI suggest they’re building core tech here, not just support teams. Salaries up to $440K signal the war for talent is real. Meanwhile, UW’s protein design lab (led by a Nobelist) is using AI to create life-saving medicines. From antivenoms to next-gen antibiotics, the Pacific Northwest is pushing AI’s boundaries in both silicon and biology. Are we seeing the rise of a new AI corridor? Here’s hoping proximity leads to more collaboration than corporate rivalry. What’s your take is this clustering good for innovation, or just hype? #AI #SeattleTech #MachineLearning #Innovation #Biotech

9

Our team at Meta just launched SAM 3 and SAM 3D - the next generation of Segment Anything Models! Check out these models in our demo playground: https://lnkd.in/eymcBgBa SAM 3 - Builds on SAM 2 with two major upgrades: text prompting and visual example prompting! SAM 3D - Reconstruction of objects and people into the 3rd dimension from a single 2D image! Learn more: SAM 3: https://go.meta.me/d74ec5  SAM 3D: https://go.meta.me/f019f3

54

4 Comments

I’m excited to share my latest blog post on how we built multi-armed bandit framework to uplevel our experimentation platform and accelerate experimentation. 🚀 Read more on the DoorDash Engineering blog: https://lnkd.in/gcp4_6QY

50

2 Comments