I build AI-powered applications and infrastructure. Most of my work involves autonomous agents, LLM tooling, and shipping real products — from computer vision systems to full-stack web apps. The dashboard above tracks my AI pair programming workflow in real time.
| Project | Description |
|---|---|
| 🪵 Longhouse | Centralized platform for managing AI agent sessions, insights, memory, and orchestration |
| ⚡ LLM Benchmarks | Benchmarking LLM inference speeds across providers — 13 stars |
| 🐣 Hatch | CLI tool for spawning headless AI agents (Claude, Codex, Gemini) |
| 🔧 MCP Tools | Lightweight MCP server facades for Claude Code — 90%+ token reduction |
| 🎁 Code Wrapped | Spotify Wrapped for your AI pair programming year |
| 📜 Agentlog | Canonical parser for AI agent session logs (Claude, Codex, Gemini, Cursor) |
| Project | Description |
|---|---|
| 🛑 Stop Sign Nanny | AI + IP camera system that tracks and scores vehicle behavior at intersections |
| 🌅 HDR Pop | Transform standard photos into HDR using AI-powered gain map generation |
| 🏀 March Madness LLM | NCAA bracket simulator with AI-powered decisions and a React frontend |
| 🍷 This Wine Does Not Exist | Generating fake wines with GPT-2 + StyleGAN |
| 🌊 FloodMap USA | Interactive flood risk mapping with elevation data |
| 🤖 AI Tools Directory | AI agents that discover, catalog, and organize emerging AI tools |
| 🖥️ Pixel Pilot | AI agent for completing computer tasks via screen control |
| Project | Description |
|---|---|
| 🚗 MPC Vehicle Controller | Model predictive control + computer vision for autonomous vehicle steering — 21 stars |
| 🦾 Robotic Control with DRL | Deep reinforcement learning for robotic control in Unity |
| 🎯 PID Control | PID vehicle controller for autonomous driving |
| 🛣️ Lane Tracking | Image processing pipeline for autonomous lane detection |
{{SHIPPING_TABLE}}
The dashboard at the top updates automatically every 6 hours. It tracks my AI-native development workflow — combining traditional git commits with AI pair programming sessions across multiple tools.
- 4 AI coding agents running in parallel: Claude Code, OpenAI Codex, Gemini, Cursor
- Avg {{AVG_TURNS}} turns/session — deep problem-solving, not quick prompts
- All metrics computed locally and via the Longhouse API
{{LANGUAGE_BADGES}}
Data sources
- Git activity: GitHub API + local
git log - Claude sessions:
~/.claude/projects/ - Codex sessions:
~/.codex/sessions/ - Cursor sessions:
state.vscdb(SQLite) - Gemini sessions:
~/.gemini/tmp/*/logs.json
Detailed Stats (Last 30 Days)
- Commits: {{COMMITS_30D}}
- Languages: {{LANGUAGES_30D}}
- AI Sessions: Claude {{CLAUDE_30D}} · Codex {{CODEX_30D}} · Cursor {{CURSOR_30D}} · Gemini {{GEMINI_30D}}
- Total Turns: {{TURNS_30D}}
Last updated: {{UPDATED_AT}}