In progress
Eval harness starter (reference build)
An open-source starting point for evaluating LLM features, with test sets, graders, regression checks in CI and a simple report. Built in the open alongside the writing.
- TypeScript
- Evals
- CI
Projects
Things I've built in the open
Reference AI systems I build on my own time, with the code and the reasoning behind it, plus some earlier tools.
In progress
Guarded agent (reference build)
A small tool-using agent with the production parts most demos skip, such as permissions, prompt-injection defences, human approval steps, tracing and cost limits. Deployable on AWS.
- Agents
- Guardrails
- AWS
Live
Lamb CLI
An npm CLI that builds and ships AWS Lambda code faster, with TypeScript support out of the box.
- TypeScript
- CLI
- AWS Lambda
- esbuild
Earlier work
Quick Chat
No-login chat rooms you share with a link. Real-time over GraphQL subscriptions on a fully serverless AWS backend.
- React
- GraphQL
- AppSync
- DynamoDB
Earlier work
Resume Builder
Build a resume, pick a design and export to PDF. A React front end on a serverless AWS backend.
- React
- Gatsby
- AWS Lambda
- DynamoDB