Priya Subramaniam

Non-technical users need to find, trust, and explore data without creating work for engineers.

Database-connected access control prevents both bottlenecks and security disasters.

Larger context windows hide a sharp accuracy cliff when schemas approach real-world size.

Unit tests catch metric logic errors before warehouse data exposes them.

Organize metrics in a semantic layer to stop duplicate definitions.

Centralized metrics prevent the same data from telling different stories to different teams.

Semantic layers let BI tools share one definition of every metric.

Self-service data access requires thoughtful permission design, not just database credentials.

Query tracing reveals why SQL is slow; pipeline tools only catch that it broke.

Capture who ran production queries—not just that they ran—to survive compliance audits.