More details coming soon!

More details coming soon!

More details coming soon!

This talk presents our hands-on journey of building a data lakehouse for a mid-sized manufacturing company.

Starting from a heterogeneous analytics landscape consisting of ERP, MES, CRM, end-of-line test data, and a legacy data warehouse, our vision was to consolidate scattered data and business logic into a single central platform.

Based on the resulting technical and business requirements, we derived an end-to-end data architecture and identified Starburst Enterprise as a strong fit for many of the required capabilities.

This talk covers our evaluation approach, the resulting architecture, our gradual migration path, and the lessons learned along the way – including why implementation progressed significantly faster than initially expected.

Key takeaways:

• Technical validation before implementation: A criteria-based evaluation covering 140+ aspects and hands-on PoCs reduced uncertainty before committing to the architecture and supported management decision-making.
• A best-of-breed architecture in practice: Starburst Enterprise for federation and governance, S3-compatible object storage, and dbt as the central transformation layer.
• Design for heterogeneous interfaces: Various source systems require different integration patterns. Combining native Starburst connectors with a custom-built Python library enabling data ingestion via RFC, REST, ODBC, ADLSv2 and JDBC.
• Gradual migration instead of big bang: Read/write access to the legacy data warehouse from the Lakehouse keeps existing data models and dashboards operational during transition.
  Demand-driven rollout creates immediate value: Developing features together with business domain teams and prioritizing concrete use cases enabled us to deliver value early while expanding the platform step by step.