Amazon's PX Central Science team managed hundreds of datasets across 12 industry verticals. Metadata was manually written, quality checks were ad-hoc, and PII obfuscation required custom scripts per dataset — creating bottlenecks that slowed data sharing, introduced compliance risks, and consumed 60% of the team's bandwidth on routine data management tasks.

DataOps Suite automates the entire data lifecycle: a 4-step Ingest Wizard handles upload, quality validation, metadata generation, and catalog publication. The Data Catalog provides searchable, browsable access with full schema documentation, quality reports, and lineage tracking. The Obfuscation Service applies format-preserving HMAC-SHA256 transformations to PII columns with deterministic, reversible results — enabling safe data sharing while maintaining referential integrity.

The Ingest Wizard accepts CSV/JSON uploads (capped at 20 rows for LLM cost control), runs a hybrid quality pipeline (deterministic + LLM semantic checks), and generates column-level metadata via Anthropic Claude. Published datasets are persisted to the catalog with full provenance tracking.

The Obfuscation Service uses HMAC-SHA256 via the Web Crypto API (crypto.subtle.sign) with a deterministic seed (DEMO_SEED_2024). In production, the seed would be a KMS Customer Master Key (CMK) with IAM-controlled access — the seed IS the re-identification key, so access control is the security boundary.

Format preservation ensures obfuscated data maintains the same structure as the original: emails remain valid email formats, phone numbers keep their digit patterns, SSNs preserve the XXX-XX-XXXX format, and names remain pronounceable strings. Running the same input twice always produces the same output (deterministic), enabling cross-dataset joins on obfuscated keys.