Physician by training,
builder by instinct —
figuring out where AI
and medicine go next.

I trained and practised as a physician, then spent years in research. My doctoral work at RWTH Aachen — graded summa cum laude (the highest distinction) — developed a method around antipsychotic-induced weight gain, and is published in the peer-reviewed literature.

I defended that thesis with an AI research system I built myself — a second brain that turned analyses that used to take me weeks into minutes. That experience was the eye-opener: I shouldn't be the only physician with access to that kind of leverage.

This is the lens I see everything through. Medicine taught me to care about evidence, edge cases, and the gap between what looks promising and what actually helps a real person. That instinct is exactly what I bring to AI.

Right now I help companies put AI to work in ways that actually pay off — finding the high-leverage automations, then building them. With Bobshop, a European retailer, that has meant turning everyday AI tools into real operating leverage rather than novelty.

Lately the work has converged on one discipline: evaluation. Most AI agent projects die in production — not because the model is weak, but because nobody can prove the system works. At Bobshop I built that proof: an eval pipeline that graded 248 production outputs against a six-dimension rubric, LLM-as-judge, traced in Arize Phoenix — failure modes named, traced, and fixed.

I think of this as honing the craft. Every company I help is a step closer to the thing I really want to build: AI that meets people where medicine and health actually live. The practical work and the bigger goal aren't separate — one funds and sharpens the other.

I weigh things carefully before I commit — and what draws me in is people and problems that are genuinely interesting and push something forward. If one of these sounds like you, that's reason enough to talk.