Translating Nubank’s Data Lake Access Control Model Into Databricks Unity Catalog

Nubank implemented consistent, fine-grained access control across its multi-platform data environment by translating its internal data lake access control model into Databricks Unity Catalog permissions. This model uses centrally managed "Permission Groups" (based on geography, data domains and column sensitivity) to govern access to over 300K tables for 5K+ users, enforcing control at the table and column level. The presentation will detail the architectural implementation, the programmatic translation process using Unity Catalog APIs, and the application of read-time masking for sensitive columns. It will address practical challenges in modeling fine-grained permissions and operating at scale, providing attendees with a reusable architectural pattern and concrete lessons for designing scalable, enterprise-grade access control on Databricks.