Grab's engineering team details the three foundational platforms powering their data mesh (Signals Marketplace): Hubble, a metadata catalog built on DataHub with an event-driven certification engine; Genchi, a self-service data quality observability platform; and the Data Contract Registry, which formalizes producer-consumer agreements. Hubble computes certification states (Uncertified, Certified, CertifiedPlus, Revoked) automatically as metadata changes. Genchi enforces freshness, completeness, schema stability, and semantic validity checks, with a Sync with Pipeline feature that ties test execution to actual Airflow pipeline completions. The Data Contract Registry stores versioned JSON contracts referencing enforceable quality rules and routes stakeholder notifications on contract changes. Together, these tools reduced the number of heavily-used P80 datasets by over 58% in one year, creating a trustworthy, AI-ready data marketplace.
Table of contents
IntroductionHubble: The data discovery and governance layerGenchi: The data quality observability layerData Contract Registry: The producer–consumer agreement layerConclusionWhat’s nextJoin usSort: