I build full-stack AI products.
Multi-agent workflows end to end — orchestration, evals, streaming UIs. Currently shipping Margin AI, a copilot that catches what clients mean on live calls.
Enough talk.
Let me show you what I’ve built.
Curated Intelligence
GIS used to need a specialist. We replaced them.
Five AI agents in parallel — profiling, locating, scanning, analyzing, plotting — to answer in seconds what used to take a GIS specialist days.
Ask the agents…
Fast only matters when the product still feels real.
I move like a founder, but I build like the thing has to survive real users. That is the difference.
MVPs in days
I cut through the vague parts early, then ship the smallest real version people can actually use.
Deep research first
Before code, I map users, market, stack, and failure paths so the build does not drift halfway.
Product taste
I obsess over where a screen feels confusing, slow, fake, or too clever, then make it feel obvious.
AI-accelerated full stack
Web, APIs, data, infra — and native macOS desktop apps. I own the whole stack and ship the product you imagine, end to end.
Tools I trust to ship real products.
The AI I build with, the AI I ship into products, and the foundation it all runs on.
AI-native, end to end
Coding agents
the AI I build with
AI APIs & SDKs
what I ship into products
Core engineering
the foundation it runs on