WatchTower / live app / source / architecture
GridMind / live / source PolicyLens / live / source (team project, I built the RAG engine)
CaseCompass / live / source TidyFlow / download / source
Transformer / source PennyPilot / try the bot / source
More builds
| Project | What it does |
|---|---|
| LLM Semantic Cache | Caching layer for LLM calls on Redis and Jina embeddings. Similar questions come back in milliseconds instead of seconds. |
| Klypse | Chrome extension that answers questions about any YouTube video, built on FastAPI, LangChain, FAISS and Groq. |
| NetLink P2P | Terminal messenger in pure C++ over raw TCP sockets and POSIX threads. |
| ZombieKill | Drag a locked file onto it and it kills whatever process is holding it. C++. |
| LocalSqueeze | Shrinks photos and videos to an exact target size, fully offline. |
| Real Estate AI | Price prediction (R² 0.92), analytics and a recommender. Live on Streamlit. |
| Fraud Detector | Model that flags fraudulent transactions. |
building: WatchTower # web-watching agents that only ping you when it matters
contributing: [prefect, livekit-agents, pipecat, txtai, deer-flow]
exploring: [distributed systems, agent orchestration, RAG evaluation]
ask_me_about: [RAG that cites sources, FastAPI at scale, Redis queues, LangGraph]
believes: "build things that work in the real world, not just in notebooks"
