About me

What I do

I help engineering teams build AI they can actually understand and improve. I get into your codebase, look at your data, and build evaluation systems that turn "it feels broken" into "here's exactly what's wrong and how to fix it."

At Frigade, we achieved 40% better responses and 75% faster latency. At Sidecar, 35% better response quality. Real visibility into what your AI is doing, plus systematic ways to improve it.

Why evaluation matters

Most AI teams ship and pray. Eyeball a few outputs, say "looks good," hope for the best.

Then something breaks in production. A customer complains. Everyone scrambles—pulling logs, squinting at embeddings, tweaking prompts, hoping the problem disappears.

Teams that ship reliable AI do something different. They measure. They have dashboards. They know their failure modes before users find them. When something breaks, they have a hypothesis in minutes.

That's what I help teams build.

How I work

I'm not interested in writing reports that sit in drawers. I work alongside your team, building real systems with real code. The goal: make myself unnecessary. By the time we're done, your team owns everything and understands it completely.

Three ways we can work together:

• Advisory ($5k/month) — Weekly calls, email access, strategic guidance. You execute, I advise.
• Embedded partnership (from $20k) — I join your team for 6-8 weeks. We build together. I transfer everything I know.
• Done for you (from $40k) — You're slammed and need results fast. I build it, document it, hand it off.

See full details on Services page →