Stock Intelligence Platform
A store stock system that grew up in public: it started as a Python desktop tool doing counts, sales and waste from CSV exports, and evolved into a four-layer production platform — React front end, Express API, dedicated service layer, MySQL under Drizzle. The interesting part is the order brain: demand forecasting over 21,000+ sales records with confidence-scored suggestions and a four-tier urgency model, built by someone who has actually run the stock counts it replaces. Performance is engineered, not hoped for: connection pooling, per-query monitoring, sub-200ms p95 responses at an 82% cache hit rate. Security the same way — multi-tier rate limiting and hardened headers are part of the middleware stack, not bolt-ons.





