Eleven years shipping products end to end.

How I work

Full stack, in the sense that I take a thing from an idea to something real people use — and then keep it running.

1. 01

   AI in products, not just in the editor

   I put model providers into things real people use, and then do the unglamorous half — evaluating and monitoring what comes back once users are involved.

   • Anthropic
   • OpenAI
   • Vertex AI
   • Gemini
   • Perplexity
   • Replicate

   At Prosper: a real-time speech pipeline on Chirp v3 and Vertex, and an LLM evaluation framework in Python with Pydantic, LangChain and Langfuse, prompt-chaining across four competency pathways.

2. 02

   AI tooling every day

   Claude Code, Codex and Copilot are part of how I work, not an experiment I ran once.

   • Claude Code
   • OpenAI
   • Copilot

   It's why one person can keep fourteen products in production across web and two app stores — and why I have opinions about where these tools help, where they quietly cost you, and how to review what they produce.

3. 03

   Equally at home on web and mobile

   Next.js on web, React Native and Expo on mobile, sharing a backend through API routes and server actions. Most of my work has been web; the mobile half is the unusual part.

   • Next.js
   • React
   • TypeScript
   • React Native
   • Node.js
   • PostgreSQL

   On web: a PHP MVP rebuilt on Next.js and Server Components at Yarn, 40% faster; a conversion funnel taken from 17% to 30% at Lendable; 15+ features on BPP’s Apex platform. On mobile: eight years of React Native across Lendable, Yarn, MDRx and Aula, plus ten of my own apps through App Store review.