About Paul

I kept crossing boundaries because the interesting problems live between them. I make AI products for people who work in complex domains: AI-driven drug discovery, clinical trials, autonomous vehicles. Curing cancer. Saving babies.

I have a PhD in Human-Computer Interaction, a decade in genomics and proteomics, and 18+ years shipping production code. That combination means I can sit with a drug discovery scientist, understand their workflow, design the interface, and build the system behind it. I set product roadmaps, run research studies, write production code in React and Python, evaluate foundation models, and ship. My prototypes are production-ready, not mockups, which is why my designs ship 85% faster than typical design-to-engineering handoffs.

The through-line across all of it is a question I've been chasing since grad school: why are experts skeptical of computational tools and AI, and what does it take to earn their trust? I've pursued it across four years pioneering agentic AI for drug discovery and autonomous vehicles at AWS, a decade shipping CE-IVD certified clinical decision support at Roche for physicians making life-or-death treatment decisions, and a PhD that gave me the research vocabulary for what I'd been observing all along. The hardest problem in AI isn't making it powerful. It's making it trustworthy—building systems where users understand what the AI is doing, can intervene when needed, and don't lose the parts of their work that make it meaningful.

Outside of work, I'm usually building something; AI tools, self-built solar arrays, Powerwalls, an electric '62 Vespa I converted myself, and a house I built from the ground up. But probably fixing something on that house.