PII-Aware Feature Store
Governance-first feature management for regulated AI systems
Overview
This project is a feature store designed for regulated environments where personal and sensitive data must be tightly governed. The system treats data governance as a core primitive rather than an afterthought.
PII Classification
- Implemented explicit PII classification at the individual feature level
- Distinguished between raw PII, derived PII, and non-sensitive features
- Enforced classification during feature definition and publication
Access Control & Masking
- Enforced access controls based on consumer identity, role, and use case
- Applied dynamic masking and redaction rules for sensitive features
- Prevented accidental feature leakage into unauthorized ML workloads
Auditability & Compliance
- Logged all feature access events with full metadata for compliance review
- Provided traceability from model inputs back to source feature definitions
- Designed to support regulatory audits and internal governance reviews