01 / CONTEXT
The problem to solve
Analysts need answers to retention questions, but raw SQL and warehouse schemas are not suitable interfaces for every business stakeholder.
02 / ENGINEERING
From need to system
- Connect an application client and server through the Databricks AppKit structure.
- Use SQL Warehouse queries and AI/BI Genie as the data-interaction layer.
- Separate business-facing questions from platform-specific implementation details.
- Package the solution with a Databricks bundle configuration.
03 / DESIGN DECISION
A meaningful choice
Make governed data accessible through an application interface while keeping the warehouse and semantic layer explicit.
04 / EVIDENCE AND LIMITATIONS
What is actually documented
The private repository includes React/TypeScript client and server code, AppKit plugins, and Databricks deployment configuration. It does not prove a live customer rollout or retention lift.
05 / NEXT VALIDATIONS
How I would strengthen the proof
- Demonstrate a fully traceable question-to-query-to-answer flow.
- Validate row-level permissions and ambiguous questions.
- Measure usability against a SQL-only baseline.
These are planned validation steps, not already measured results.