I build AI products, local-first developer tools, and experimental systems that make powerful models easier to run, test, and ship.
- AI app builders — agents, sandboxes, streaming progress, and persistent project state
- Local LLM tooling — Rust workspaces, OpenAI-compatible APIs, quantization, profiling, and Python bindings
- Self-learning agents — coding systems that generate, evaluate, critique, and improve
- Performance experiments — model inference, benchmarks, memory optimization, and hardware-aware runtimes
| Project | What it is | Stack |
|---|---|---|
| oxidize | Rust workspace for local LLM tooling: CLI, server, quantization utilities, core model primitives, and Python bindings | Rust, pyo3, WASM |
| zapdev | AI-powered app builder with live sandboxes, streaming agents, and persistent project state | Next.js, React, Convex, Inngest |
| self-learning-ai | Locally-run self-improving AI coding agent with sandbox evaluation and LoRA learning loops | Python, ML/AI |
I'm especially interested in projects that combine:
- Fast local AI — practical inference on real hardware
- Agentic coding loops — generate, run, inspect, improve
- Beautiful product surfaces — tools that feel polished, not just powerful
- Open systems — composable APIs, CLIs, and self-hostable infrastructure
- GitHub org: Zapdev-labs
- Website: zapdev.link
- Best place to explore my work: check the repos and pinned projects
building AI products, one experiment at a time