Evaluation Engine

Monitoring tells you something broke. Gating stops it from spreading.

The Evaluation Engine runs configurable quality checks at three levels — table (row counts, schema, partition health), column (null rates, distinct counts, ranges, patterns), and row (duplicates, referential integrity, business rules) — then decides whether a dataset is allowed to promote.

Everything is expressed as SQL and executed in Spark, so it scales with the data rather than against it.