How I Work
I don't use AI
to skip steps.
I use it to move faster through all of them. Real research. Validated concepts. Production code. No handoffs, no translation loss.
These aren't steps — they're the things that matter in every project, in whatever order the work demands. The tools keep changing. The discipline doesn't.
Real users. Synthesized at scale.
Research before requirements. That means talking to actual users — observing how they behave, not just what they say. Synthesizing across interviews, usability sessions, and behavioral data to find the real pattern. The output isn't a deck. It's a sharp problem definition with evidence behind it.
Diverge fast. Converge with evidence.
I explore more directions than I'll ever present — fast, rough, and wide first. Usability thinking isn't a phase that comes later. It's in the first sketch: how will users navigate this, what mental models do they bring, where will friction appear? I make decisions based on real comparison, not the first good idea.
No black-box handoff.
I don't stop at the spec. Using AI as my coding co-pilot, I generate and refine frontend code directly in the engineering repo — then engineers review and ship. I'm not a developer, but I'm close enough to that line that the translation gap between design and code gets closed, not papered over.
The result?
Interfaces that ship faster, break less, and stay closer to user needs than anything designed in a silo and handed off at the end.