A multi-API pipeline that compresses a ~5-hour open-source research-and-drafting process into ~5 minutes: OSINT gathering → AI synthesis into a structured script → matched public-domain visuals, handed to a human creator to voice and finish.
An educational engineering experiment wiring a local Claude model to a broker's official Model Context Protocol server, with a rules engine that reacts to market-data signals — built and run against a simulated (paper) environment.
Clients hire an automation partner for discipline — secure, enterprise-grade systems that ship and run. Labs is where I pressure-test new protocols before they earn a place in that work. Every experiment here carries a clear scope-and-compliance note; none of it is a product or advice.