Regulated Data Engineering at Scale

Built GDPR-compliant data pipelines processing hundreds of millions of records a day for a European bank. The operating principle: PII handling belongs at ingest, not at point of use. If personal data enters a pipeline unclassified, every downstream system inherits the compliance problem. Classify, mask, or tokenize at the boundary — not later, and not optionally.

The connection to AI work is direct: the same discipline is what makes an AI system deployable in a regulated environment at all. Knowing exactly where personal data is, who can see it, and being able to prove that after the fact isn’t a nice-to-have around a good model — it’s the gate a model has to pass through before it’s allowed near real data.