Retail Intelligence Platform

Overview

A full-stack sales & inventory intelligence platform for a 3-store grocery chain, built on a year of daily sales and inventory history. It answers the questions an owner actually asks: which products matter, which are quietly tying up cash, what to reorder and when, and where sales are behaving abnormally. A React dashboard sits on a FastAPI backend with JWT auth, Redis-backed caching and a live WebSocket feed — record a sale and every open dashboard instantly sees the stock move and any low-stock alert.

What we did

• FastAPI backend with JWT auth (admin roles), SQLAlchemy models and Redis-backed response caching.
• OLS demand forecasting with day-of-week seasonality and a 95% prediction interval, plus train/test backtesting.
• Automated reorder engine — safety stock and reorder points from forecast mean, variance and supplier lead time.
• ABC/Pareto revenue classification, gross-margin contribution and inventory-turnover analytics.
• Sales anomaly detection and a 30-day/60-day store-performance comparison normalised for scale.
• Real-time low-stock alerts pushed to every connected dashboard over WebSocket on the sale write-path — no polling.

Tech stack