The arc.
I started in tech before most people had broadband at home. Over 25 years I've been a developer, a consultant, a startup founder, a strategist, and now — an AI architect. The through-line has always been the same: systems that work in the real world, not just the lab.
A granted AI patent sits somewhere in that arc. So do two startups I built from scratch. So does a lot of failure. All three taught me roughly the same thing: execution beats ideas, every time.
How I think
Execution beats ideas, every time.
Every great architecture I've shipped was the second or third version. The first one was always wrong — and the gap between knowing that and accepting it is where most teams stall.
Most AI projects fail in the architecture, not the model.
Swap the model and the same project still dies. The failure was in the data path, the eval harness, the cost model, or the human-in-the-loop boundary — months before the demo.
The best engineers I know are obsessively specific.
Vague systems thinking is a tell. Precision about inputs, outputs, failure modes, and SLAs is what separates a working system from a slide deck.
Complexity is a debt you always pay back.
Every framework, every abstraction, every clever pattern compounds. The strongest systems I've built had fewer moving parts than the ones I was tempted to ship.
Technical depth
I work across the full agentic stack — orchestration frameworks, routing and caching layers, model serving, eval harnesses, and the observability tissue that holds it together.
I read papers I disagree with on purpose. I stress-test new frameworks the week they ship — not to chase novelty, but to know where they break before a client does. The blog is where most of that thinking eventually lands.
Outside work
Driven by curiosity, family, and the conviction that the most interesting problems in AI right now aren't in research labs — they're in the gap between what works in a demo and what survives its first month in production.