Problem
Retail transaction data (84.51°/Kroger-style) holds churn, CLV, and cross-sell signal that stakeholders can't reach without an interactive, deployable analytics product.
Architecture
FastAPI backend over Azure SQL with ML models for CLV, churn, and basket prediction; React + Vite dashboard with household-level search, CSV upload/refresh, and a persistent SQL-backed dashboard cache for stable Azure deployment.
Contributions
- Deployed the platform end to end: auth, data storage, ML endpoints, interactive reporting, and the dashboard cache.