Engineering Notes

Writing about the work behind secure, useful software.

Practical posts on architecture, AI integrations, privacy workflows, and lessons learned shipping software in real environments.

First Post PII Detection Tokenization AI Support Workflows

Adding Reversible Tokenization to a Presidio Pipeline with Protegrity AI Developer Edition

A migration story about moving beyond redaction when AI systems still need structurally valid, reversible, policy-controlled protected data.

Key Takeaways
Redaction broke context

The model could no longer distinguish repeat customers or preserve cross-ticket signal.

Tokenization restored utility

Protected values stayed structurally valid and consistent enough for downstream AI use.

The abstraction paid off

A clean processor boundary kept the migration contained to a few files.

More posts can drop into this page as simple cards without changing the overall structure.