I build AI products that inspire real users.

AI products fail when teams treat models like features instead of systems. A feature ships and works. A system ships and degrades, hallucinates, and surprises you at 3am. The product discipline for AI isn't just roadmap management — it's building the evaluation infrastructure that tells you when your system is lying.

My method: Discovery → System design → Experimentation → Evaluation → Iteration. I focus on products where model capability is core, not cosmetic — where the intelligence layer is the product, not a chatbot bolted onto a CRUD app.

My background in Human-Computer Interaction means I design for the human failure modes, not just the model ones. Five years of shipping across distributed systems, crypto infrastructure, and developer tooling taught me that the messy parts — edge cases, error states, confused users — are where product quality actually lives.