I build AI agents that do actual jobs, not demos. My pipelines forecast warehouse demand, raise purchase orders, flag anomalies in CRM data, and at one point detected plankton. Hackathons take you to strange places.
Right now: shipping CrowdStacks, a crowdfunding dApp on the Stacks blockchain. On-chain escrow written in Clarity, goal-based withdrawals, self-serve refunds when a campaign misses its goal. Built as a two-person sprint with real PRs and real code review.
| Project | The pitch |
|---|---|
| chainly | Three agents run a warehouse: one watches inventory, one forecasts 30 days of demand, one writes the purchase orders. LangGraph, Groq, FastAPI, deployed on AWS. |
| orchestrix | An 11-node LangGraph pipeline that ingests CRM data, hunts anomalies, forecasts KPIs, then triggers n8n automations so humans only see the interesting parts. |
| Grizzly | An open editor-in-a-skill for fiction writers. It diagnoses your draft, tracks structure, and remembers your story across sessions. The author keeps the pen. |
| realtime-chat-app | WebSocket rooms with live presence. My most starred repo, which proves people just want to talk. |
| diatoma | Ensemble ResNet models scanning images and video for plankton. Built for Smart India Hackathon 2024. |
| ledger | Tamper-proof CCTV evidence. Every video gets a SHA-256 hash logged on Ethereum/Polygon, so nobody can quietly edit the footage. |
| gesture-enabled-presentation | Drive PowerPoint with your hands via OpenCV and MediaPipe. Clickers are for cowards. |
Daily drivers: Python, TypeScript, FastAPI, LangGraph, React/Next.js.
Also in the toolbox: TensorFlow, scikit-learn, OpenCV, Node.js, Docker, MySQL, MongoDB, AWS (EC2, Elastic Beanstalk). And yes, Clarity smart contracts, which I did not expect either.
