I'm a data scientist and engineer in Atlanta, working on enterprise data governance: the layer of a data platform that decides whether a dataset is owned, registered, described, tested, and properly entitled before anyone builds on it.
I build the tooling that makes those decisions repeatable. Compliance-check pipelines that turn written policy into per-dataset pass/fail results, with new domains added as configuration rather than code. Certification workflows that render those results into a signed certificate and walk it through review with one-click, tamper-evident links. Governance knowledge graphs that keep an assistant's understanding of the policy wiki current, where every routing decision can be traced back to a page and a reason. And agent skills on top of all of it, gated by evaluation sets so a change that regresses an answer never merges.
The way I think about it: governance tooling should be thoughtful rather than loud. The best control is one that catches an issue by design, before a human has to notice it, and explains itself well enough that the human can trust it. Deterministic first, models where they earn their place, and a test that says what "correct" means.
Pattern code available on request.