Data: The Land DevOps Forgot
summary
Nubank moves 50 petabytes through 300,000 nightly jobs and 7 million lines of Scala for 100 million customers, and replaces monolithic ETL that runs 14+ hours with a domain-driven data mesh backed by Spark, Iceberg tables, a commercial data catalog and AI regex classification. A self-service platform adds observability, guard rails, SLOs, lineage and cost visibility; CI build time falls from six hours to ~3.5 minutes and teams gain autonomous deployment and measurable quality improvements, rollout showing promising results.
AI-written summary · about
chapters (10)
- 0:00Intro
- 2:45Operational vs analytic
- 7:28The opposite of DevOps
- 8:07New paradigm required
- 9:35Data Mesh
- 18:50Supporting the data owner
- 22:10Platforms create incentives
- 28:55Make the need visceral
- 31:17Summary
- 33:07Outro
related talks
32mNubank's experience scaling microservices operations
54mData Mesh in Practice
26mDataOps: Accelerating Digital Transformation
44mBreaking a Monolithic Devops
01h 00mData Mesh: Challenges and Opportunities
38mData Secrets from a Platform Engineer
30mDevOps For Data Teams
30mA Devops Approach to Data: A Tale Through the Trenches in Search of a Cloud Data Revolution
