Full-stack engineer who obsesses over whether people can actually use what I build.

I led the scheduling v2 rewrite and built it alongside the clinic, embedded on-site with front desk staff. I co-designed in live workflows, iterated on real friction as it surfaced, and turned field feedback into a prioritized roadmap for product and engineering.

// Ship it while sitting next to the person using it.

I built the TipTap-based charting editor clinicians use to write, review, and print notes, including AI-generated drafts rendered from LLM markdown and assessment suggestions that a clinician accepts or rejects.

// The human stays in the loop, so the loop has to be fast.

I ran evals on generated summaries and assessment suggestions for accuracy and consistency, then turned what I learned into evals-by-example, a nine-step tutorial.

// If you can't measure it, you're just hoping it's right.

It pulls from PagerDuty, Sentry, Datadog, LogRocket, and Kubernetes logs and suggests a likely root cause, but only when its confidence clears a threshold. Patient data never shows up in its output.

// "I don't know" is a feature.

I built the shared library for Widen's application suite and led its accessibility audit. Years later I built a shadcn/ui-based library at Akido, along with front-end review standards that now run in our automated code reviews.

// Good defaults scale better than good intentions.