What I’m focused on this month.

Leading the UI for AI Screening inside LinkedIn’s Hiring Assistant. The interesting work lives in the seams — how to make the model’s confidence legible without overselling it, how to give recruiters a clean way to disagree with it, how to fail honestly when the model isn’t sure.

Reading more about evals for LLM-based product features. Lots of opinions in the space, less consensus than you’d expect.

Off the keyboard: shoulder season. Too late for skis, too early for the bike legs to feel right. Trying to fix the second part with discipline.