About
I'm Yu Zhu, mostly Nemo online. A backend / full-stack engineer based in Seattle. I write about AI and software, and build the ideas I trust into things that run, work, and live online.
By day, I've spent years shipping production systems at Amazon Ads and Microsoft — and put LLMs into real products early: a sales assistant with RAG grounding, tool calls, entity disambiguation, permission-scoped guardrails and auditable calls; a permissions service powering direct / hierarchical / transitive authorization across ~20M relationships. Earlier at Microsoft, a telemetry-driven recommender shipped globally to 217K+ seats with a statistically significant A/B uplift, and a consolidation of 6 fragmented tools into one dashboard that cut case triage time in half.
By night, I've rebuilt most of my own life and workflow on top of AI: orchestrating the everyday with headless agents, wiring semantic retrieval into my notes, weaving scattered judgments into systems — this site, a cognition renderer, an investing north star, local-GPU voice cloning… they all live in Building.
I believe the best way to understand something is to build it. So behind almost every opinion I write, there's something actually running.
Looking for
I'm looking for a place to combine production-grade AI / agent systems with solid software engineering and build real things — AI applications, agent platforms, full-stack all welcome. I bring two things that rarely come together: experience taking LLMs to production at scale (RAG, tool calls, guardrails, large-scale permissions), and the habit of building what I imagine and using it every day. If you're working on something interesting, reach out.