—

I'm a Product and Program Manager who works hands-on with AI. My edge isn't that I use the tools — it's that I architect them: choosing the model, designing the prompt strategy, defining what "good" means, and shipping the workflow around the uncertainty.

The projects here are autonomous and multi-agent systems I designed and built — most of them on the Claude Agent SDK. They run real workloads: an unattended trading operation, a quant research swarm, a market-implied geopolitical gauge. Each is opinionated about safety and auditability, with humans kept at the edge for judgment and override.