When the same person designs and builds a feature, less gets lost between intent and implementation. I build in React and TypeScript — Next.js, Astro, Gatsby and React Native depending on what the project needs — with Tailwind or MUI, Zustand for state, and real-time and streaming interfaces wired directly to the model APIs.

Everything ships accessible to WCAG and covered by tests: Jest for units, Playwright end to end. Content runs on headless CMSes — Sanity.js, Strapi, WordPress — with GROQ or GraphQL, so the people who own the words can change them without a deploy. Where a backend is part of the picture I work in FastAPI.

I work AI-first day to day — Claude Code, Cursor and Figma Make from prototype through to production code. The division is deliberate: AI accelerates implementation, and I own the decisions that need judgement — interaction behaviour, edge states, accessibility and system consistency.