Explores an AI-native data architecture called 'composable canonicals' that addresses the central data team knowledge problem. The approach creates machine-readable domain knowledge blocks (ontologies) per data source — e.g., HubSpot, Slack, Luma — and composes them into a unified business canonical layer. Each source canonical models one system in its own language with raw tables, transformations, and an ontology. The business canonical then resolves identity across systems and defines shared entities like 'Customer' once, avoiding multiple sources of truth. The workflow is implemented using dltHub for ingestion, transformation, and AI-assisted pipeline building.
Table of contents
Composable canonicals Link iconEach source is its own canonical Link iconThe business canonical composes them into one Customer Link icondo it with dltHub Link icon281 Impressions1 Comment