I run the legal and compliance function for a regulated fintech operating a FINRA broker-dealer, an SEC-registered investment adviser, and a futures entity, with hundreds of thousands of brokerage accounts supervised by a compliance team you can count on one hand.
That ratio works because I build. Over the past two years I’ve designed and shipped AI systems that run the firm’s compliance testing and ongoing supervision, and a framework that lets a large language model deliver investment advice that is deterministic, reproducible, and auditable, built for supervision under both the Investment Advisers Act and FINRA’s broker-dealer rules, and previewed with regulators before launch.
I call the operating philosophy supervision by design, the industry’s compliance-in-the-loop idea taken to its architectural conclusion: agentic systems do the data-pulling, formatting, and first-pass review; principals do the judgment work. The AI is never the decision-maker of last resort, and the architecture, not a policy binder, is what makes that true.