Industry systems and shipped builds

Turn your Wikipedia rabbit hole into a beautiful, shareable map. Your actual clicked path becomes a bright animated spine; all other links sit as faint context so the graph never becomes a hairball.

• Spine-and-focus graph principle: clicked path is a thick animated edge, neighbors dim, distant nodes fade. Sidesteps the Obsidian/Roam-style hairball at 200+ nodes.
• Reads inside the map: clicking a node opens a floating burrow card anchored to the node, so the map never leaves the screen.
• Reverse-engineered from the share card inward: the artifact is designed first (Spotify Wrapped lesson), then the experience fills it.

Founding engineer on an industrial copilot platform. Multi-agent AI workflows over live SCADA, PLC, MQTT, and SQL Server data on the plant floor.

• Multi-agent workflows on per-deployment knowledge graphs (2k to 5k nodes), orchestrated through LangChain.
• Shipped 15+ client POCs end-to-end, embedded with manufacturing teams.
• Built FastAPI services + a provisioning CLI for Dockerized pipelines on AWS and Kubernetes.

Offline, citation-grounded RAG for technical documents. Refuses to answer when it can't cite a source. Deployed on a local GPU workstation at the sponsor's office.

• Citation enforcement blocks ungrounded outputs before the user sees them. That is the policy that makes the system safe on regulated content.
• Hybrid retrieval: FAISS + BM25 fused via Reciprocal Rank Fusion, then a cross-encoder reranker.
• FastAPI + SSE backend, Chainlit chat UI, evaluation harness for Recall@K, nDCG@K, MRR, and latency.

Multi-agent fact-checking pipeline with adversarial debate, source-credibility weighting, and on-chain provenance via Solana memos on devnet.

• Decomposes claims via HiSS, retrieves with Gemini + google_search, weights by MBFC source credibility (~4,000 domains).
• Pro/con debate fires only when inter-agent agreement falls below 80%. Emits one of 7 verdicts (TRUE through UNVERIFIABLE), not a boolean.
• Writes verdicts to a Solana memo for permanent provenance. Voice summary via ElevenLabs TTS.

One AI-personalized sustainability action per day, with radical transparency about the carbon cost of every inference. Built in 12 hours.

• Daily actions tailored to commute, diet, live grid carbon intensity, and weather, drawn from EPA + DEFRA emissions data and 190 curated actions.
• Chrome extension and Eco-LLM dashboard track energy (Wh), carbon (gCO2), and water (mL) per Gemini prompt. Semantic caching serves repeat queries at zero added inference cost.
• Typical carbon ROI 10,000:1. Projected 12,000 tonnes of CO2 removed per year at 100k DAU.