Harden my schema with database-level validation. Go through each table or collection and add the constraints that make bad data unwritable. Apply: NOT NULL on every field that should never be empty (emails, foreign keys, amounts, statuses); CHECK constraints for invariants (amounts non-negative, end_date after start_date, status within an allowed set of values); correct column types (money as integer cents or decimal, never float; timestamps as real timestamp types, not strings; booleans not 0/1 text); and sensible defaults (created_at defaults, empty arrays over nulls where the app expects a list). In a schemaless database, add the equivalent with schema validation rules the database enforces (MongoDB has $jsonSchema validators) rather than only validating in app code. Before adding constraints, scan existing data for violations and report them; constraints will fail against dirty data. Fix the type mistakes carefully: changing a column type on live data needs a migration plan, which you should write for me rather than winging it. Deliver: constraints added per table, the data-quality violations found in existing rows, the type-fix migrations, and verification: attempt to write invalid rows directly (null email, negative amount, bad status) and confirm the database refuses them.