AI Powered Inventory Management System

Jan 2024 – Feb 2024 (10-week CSE capstone) · Active

Overview

A full-stack inventory platform for products, suppliers, stock levels, and purchase flows—designed as a 10-week final-year CSE project. Phase 1 ships production-ready foundations: JWT authentication, role-based access (admin / manager), product & supplier CRUD with search and filters, a dashboard with inventory insights, low-stock alerts, and a PostgreSQL schema managed with Prisma. Later phases add Socket.io chat, OpenAI GPT‑4 for natural-language ops, Google Document AI for invoice OCR, and demand / analytics UI.

The public codebase lives on GitHub: Milanirosefrancis/AI-powered-Inventory-Management.

What it solves

  • Single source of truth for stock—replacing scattered spreadsheets with audited CRUD, low-stock signals, and structured supplier data (including GST-oriented fields).
  • Clear split of concerns: Vite + React UI, Express API, Prisma data access, Redis for sessions/caching, and optional AI sidecars (ai-services) for GPT‑4 and Document AI.
  • Operability: Docker Compose for Postgres + Redis, seed data for demos, structured logging (Winston), and CI via GitHub Actions.

Highlights

  • Dashboards and Recharts-driven analytics views for at-a-glance inventory health.
  • Zustand for predictable client state; React Hot Toast for feedback.
  • Extensible roadmap: real-time chat commands, conversational reorder, invoice ingestion, and forecasting—without rewriting the core domain model.

Architecture

High-level view of how the React client, Express API, data stores, and phased AI services fit together.

Tech stack

React 18 + TypeScript
Vite
Tailwind CSS
Zustand
Recharts
React Hot Toast
Socket.io
Node.js + Express
JWT · bcrypt · Winston
PostgreSQL
Prisma ORM
Redis
OpenAI GPT-4 API
Google Document AI
Docker · Docker Compose
GitHub Actions