ML · Full-Stack · Case study

StockSense AI

ML-powered demand-forecasting and stock-optimization platform: safety stock, reorder points, ABC classification and EOQ optimization with interactive dashboards.

Team: Solo
Timeline: February to April 2026
Commits: 8 of 8 + 2 of 2 (two repositories)
Stack: Python · FastAPI · LightGBM · Prophet · PostgreSQL · React · Streamlit · Docker

Situation

Retailers lose money at both ends of inventory: overstocking ties up cash, while stockouts lose sales. Manual reorder rules can not keep up with seasonal, item-level demand.

Task

Forecast demand per item and turn those forecasts into concrete inventory decisions: how much safety stock to hold, when to reorder, and which items matter most.

Action

I built a forecasting service with LightGBM and Prophet for item-level time-series prediction, then layered classic inventory science on top: safety stock, reorder points, EOQ and ABC classification. A FastAPI backend serves the models and a React + Streamlit dashboard lets users explore forecasts and optimization output interactively. The stack is Dockerized for deployment.

Result

A decision-support platform that converts raw sales history into actionable stocking policy, surfaced through interactive dashboards.

2 models

LightGBM + Prophet

EOQ / ABC

optimization

Interactive

forecast dashboards