Designing with craft, building with code.

I start by asking what could go wrong. In AI, failure modes matter more than happy paths—so I design for uncertainty first, then work backwards to the ideal experience. This means showing users what the system is doing, letting them intervene, and making recovery painless.

I prototype in code when the interaction can't be faked in Figma. Timing, state transitions, and edge cases only reveal themselves when you build the real thing. My indie apps exist partly because I needed to understand problems that static mocks couldn't answer.

I studied computer science before design. That lets me have real conversations about feasibility—not just what we want to build, but what we can build well. My strongest work happens when I'm embedded with engineering and PM, not handing off specs from a distance.