Hand me the problem, not the spec.
You have something slow, wrong, expensive, or stuck, and it needs someone who can own it end to end. That is the work I want.
The parts I take are the ones nobody has cracked and AI alone won't fix: search and relevance, data at volume, AI features that have to hold up under real traffic. I go from the data all the way to what your customer touches.
What I take on
Search that returns the right thing
Relevance, ranking, and autosuggest. I've built it at Bing scale and at startup scale, and I'm deep in Elasticsearch: query DSLs, scoring, and the performance work that keeps results fast when traffic is real.
AI features that survive production
RAG, chat, agents, and MCP servers, built so your own engineers can extend them after I leave. I'm building AI systems from the ground up in my current role, past the demo and into something people depend on.
Systems that stop being slow
Indexing pipelines, query latency, data volume. I find the actual bottleneck instead of adding machines, and I leave the measurements behind so you can prove it stayed fixed.
The problem nobody has scoped yet
Stack decisions, codebase assessments for acquisitions, and the whole build when there is no spec. You describe the outcome; I come back with a working system and a straight answer about what it cost.
Why me
I don't need a spec, a ticket queue, or someone to unblock me. I work out what the real problem is, decide what to build, build it, and tell you plainly what I traded away. I move fast, sometimes overnight.
I've done this inside large companies and as the only engineer on a project. I can lead a team when the work needs one, and work across time zones when it doesn't. I've trained over 100 developers and mentored engineers one on one, so what I build doesn't need me to keep running.
Some of my work
Search
Six years on MSN Search, Live Search, and Bing, building relevance systems, data pipelines, and ML classifiers. A patent on intent detection.
AI
Led engineering for Gemini Code Assist at Google, the AI coding assistant for VS Code and IntelliJ. Built for developers who keep it open all day, where a wrong answer costs more than a slow one.
Disease modeling
I worked for Bill Gates at the Institute for Disease Modeling, now part of the Gates Foundation. I built the visualizations for its epidemiological simulations and geospatial modeling, and they were in his TED Talk "The next outbreak? We're not ready."
Consulting and startups
Past clients include Hims & Hers, Sensei Ag (founded by Larry Ellison), and CopBot, plus a long tail of shorter engagements. Brought in to run engineering at T-Mobile, I hired 25 engineers in six weeks and took on four product teams. I've founded startups of my own and been CTO at two others.
Day to day I'm in TypeScript and Python. The language is rarely the hard part, so I work in whatever the codebase is already in.