
We got llms at home.
Real language model checkpoints trained locally, then loaded and run in the browser with a custom TypeScript and WebGPU inference runtime.
Products and experiments in browser-based machine learning, applied AI, developer tools, and data systems.

Real language model checkpoints trained locally, then loaded and run in the browser with a custom TypeScript and WebGPU inference runtime.

A focused writing environment where quiet inline suggestions help you find the next sentence without handing the page over to AI.

An on-demand Next.js application that streams generated Markdown into safe, cached pages with per-page regeneration and visible cache behavior.


A motorsport analytics project that ranks races and drivers using historical Formula 1 data and a purpose-built Elo model.
