AI-native product · design engineering
I figure out the right thing to build — then build it.
Fifteen years of UX depth, now pointed at building working software — not specs. I prototype AI-native interfaces directly in production React and TypeScript with Cursor and Claude Code. The UX background is the moat: it’s why AI-assisted building produces the right thing instead of one-shot slop.
When building is easy, judgment is the scarce resource.
- 01
How the media actually feels about AI.
A live index scoring daily AI-stance sentiment across 14 news outlets — shipped solo and full-stack: Next.js on Vercel, Supabase Postgres, and a Python pipeline on a 6-hour cron. I scored with Claude Haiku over VADER because VADER agreed only 62% of the time and couldn't tell that “Wins Court Order Pausing Ban” is positive for an AI company. Token cost (~$0.01–0.02/day) was a deliberate constraint.
View live - 02
Resolving merge conflicts with judgment, not just automation.
A React/TypeScript PR-review app and GitHub Action that auto-resolves merge conflicts and presents each resolution for developer approval. Token cost was a first-class design constraint that shaped the UI; cross-model confidence scoring means no model grades its own work.
View live — coming soon - 03
Prototyping in the real codebase, not in mockups.
Prototyped merge-conflict resolution directly in the React/TypeScript/Electron app with Cursor and Claude Code — much of it shipped to production, and engineers started from my branch. Built a tokenized design system with repo-backed Figma-to-code sync.
- 04
One component, ~300 designers.
Designed and contributed a reusable CloudScape design-system component that resolved a four-year-old request, adopted by 40+ teams and ~300 designers. Sole designer on Mirador, consolidating six security tools into one platform for thousands of developers.
Discover the problem. Work it out with AI. Prototype in real code.
I own the validation, the decision quality, and the functional prototypes that de-risk the build — and deliberately hand production durability and on-call to engineering. The prototype isn’t a throwaway; it’s the working recommendation engineers start from.