AI systems running in production
Each project here is a system that runs on real data every day. The pages state what the thing does, how it is put together, and what it refuses to do, at the level a technical reader can judge. Where a project has been measured against the path it replaces, the numbers are in case studies, and the measurement method is specified in AEQ.
- AI Agents, AEQ, Orchestration, R&DAgentSaaSy: The Swarm, MeasuredA live look at an enterprise AI Agent stack: harness orchestration, cross-family judging, and AEQ scoring on every run.An animated walkthrough of one metered workflow: a request enters, the harness fans it out to specialist agents in parallel, a cross-family judge reviews the result, and a pre-registered AEQ gate rules on it before anything ships.PythonAnthropic APIOpenAI APItiktokenAEQ GridCanvasR&D demoSee the build
- Document intelligence, local-firstChiefOSAn AI chief of staff for your documents. It files them, surfaces what's due, and remembers who and what appears across all of them.A swarm of eight specialized agents that classifies, files, deduplicates, and reads deadlines out of thousands of documents. Local models, local database, one drop folder.PythonPostgreSQLNext.jsReactTypeScriptSwiftOllamaMCPIn production, in the lab313 daysSee the build