I design the software dispatchers at 5am drivers mid-route homeowners setting up wi-fi analysts with 40 tabs open can’t quit out of.
ML Dispatch, WM
Design System, WM
Self-Install, AT&T
Teaching dispatchers to trust a scheduling model
WM wanted machine-learning-powered scheduling in the dispatch tool. The model was the easy part. The real question was whether the people responsible for thousands of routes would ever act on what it suggested.
$3.2M
saved in pilot year
Making dense systems learnable
Dispatch boards, Gantt trackers, tables with forty columns and no room to spare. A new hire needs to be useful in a week; a fifteen-year veteran needs every field on screen. Both of those people are right, and I love the puzzle of serving them at the same time.
Designing the human side of ML
A model's recommendation is worth exactly nothing if the person receiving it doesn't buy it. So I design for legibility, for confidence, and for the freedom to say no — which is the unglamorous part that turns clever automation into something people actually use.
Building the systems that outlast me
Design systems, governance, contribution models, design ops. I've been the only designer on an entire product suite — the only way that math works is to let the system carry the weight. Ideally you can't tell when I stepped away.
Koi Studios
Designing for Xtrace — an AI startup spanning a Chrome extension and a seriously fast LLM agent webapp — and end-to-end designer for OpenTable's ML-powered data management platform.
Both are still in flight and not public yet — but I'm always happy to walk through the thinking on a call.