A field guide to how he works
I build the data and AI systems that scaling companies run on — and the teams that keep building them.
Every company wants AI. But AI doesn’t fix broken operations — it amplifies them. The gap is almost always data, so that’s where I start. Then comes the agent layer on top, with the evidence discipline to prove it works.
How I work on any system — the same five moves, whatever the business.
- 01 See the whole system Interconnected problems, not isolated tickets.
- 02 Find the leverage point The one build that pays out in every direction.
- 03 Build the platform One investment, made to serve many needs.
- 04 Grow the people The team that keeps building after you.
- 05 Leave it running Systems and teams that outlive you.
Four systems, one method — each worked to the depth the problem needed.
- Audience Serv
Three bottlenecks were one bottleneck. I built the platform that cleared all three — and the team that ran it.
1 → 12 data team built and trained 5 of 5 moves - FPT Software
Brought into a Fortune 500 engagement that was slipping. The stack was new to me; the shape of the problem was not.
7-figure follow-on revenue this earned FPT 4 of 5 moves - Synna
As CTO, I bet the raise on a working product over a pitch deck, not a promise — and the build is what got us funded.
$120K pre-seed, secured on the strength of the build 3 of 5 moves - Harness engineering
The bottleneck moved from writing code to trusting it. So I built the machinery that makes “works” something an agent has to prove, not claim.
50 → 39 score inflation the harness caught in itself: a blind re-judge on identical evidence 4 of 5 moves
This guide was built by the same harness it profiles — a generator writes, a blind evaluator scores it against a rubric, and a gate refuses the work until it earns a pass.
How it was made →