The Loader bulk-loads de-identified, OMOP-mapped datasets into warehouse tables — typically scheduled after pipeline success.
All-or-nothing batches
Failed loads roll back the batch — researchers never see half-published schema or partial cohorts.