I build production-grade developer tooling for autonomous AI systems—especially systems that need to be observable, testable, and actually useful in real workflows.
My current obsession: combining graph-native intelligence, local-first product design, and crisp UX so teams can understand what their agents are doing and why.
| Project | What it does | Stack |
|---|---|---|
| LoRA-jit | Benchmark-first JIT adapter router with live cold_miss/warm_hit paging telemetry and VS Code visibility. |
Python FastAPI TypeScript VS Code PEFT |
| Pres (PResolution) | Autonomous PR resolution agent that reads review comments, generates code fixes with AI, and commits directly to PR branches. | TypeScript Probot LLM APIs EJS |
| arbor | Graph-native code intelligence replacing embedding-only RAG with deterministic program understanding. | Rust tree-sitter Graphs MCP |
| Agent Lens 2.0 | Local-first visual debugger for LangChain/CrewAI workflows in real time. | Python LangChain CrewAI |
| Legion-LOQ-control | Lightweight open-source control utility for Lenovo Legion/LOQ systems without Vantage bloat. | Rust Systems Desktop |
- Autonomous agent tooling
- Graph-native intelligence systems
- Local-first architecture
- Developer experience for AI workflows
Fully generated by code from live GitHub signals—auto-refreshing as your contributions evolve.
I’m currently building systems that convert code + agent telemetry into actionable feedback loops—where routing, reasoning, and reliability can be measured instead of guessed.