I build AI systems that turn messy real-world documents into reliable structured data — and agents that act on it.
Independent AI engineer focused on intelligent document processing (IDP) and applied LLM / agent systems. I take projects from "a folder of 10,000 ugly PDFs" to clean, validated, production-grade data — and from "we wish this ran itself" to an agent that does.
🗂 Intelligent Document Processing — High-accuracy extraction from hard documents: scanned forms, technical drawings, financial statements, claims. Staged pipelines (classify → extract → validate against ground truth), corpus-scale batch processing with resume + caching, delivered as Excel / CSV / JSON.
🤖 Agentic & LLM systems — Tool-using agents, local & air-gapped LLM deployments, RAG, and evaluation harnesses. I measure accuracy and cost — not vibes.
⚙️ Shipping — Prototype to a real URL: Docker, observable pipelines, clean deploys. I deliver working software, not notebooks.
| Project | What it is |
|---|---|
| axon | Self-contained edge-intelligence agent for air-gapped infra. Local LLM (Qwen3), tool-calling, single-package deploy. |
| gerberview | Browser-native Gerber / PCB viewer. Rust → WASM + WebGL, 60 fps, zero backend — nothing leaves your machine. |
| tradestation-backtest-toolkit | Automated backtesting + statistical analysis pipeline for quant research. |
Python · TypeScript · Rust · LLM APIs (Claude · Gemini · GPT) · LangChain ·
PyMuPDF / OCR · PyTorch · Docker · Azure · WASM / WebGL
Got documents you wish were data, or a workflow you wish were an agent?