they describe it as a "data transformation framework" that is supposed to remove some of the boilerplate around setting up things like batch incremental ingestion, dependencies between data transformations (e.g. table 1 depends on table 2), testing/auditing data transformations