AI Business Platform

The system that runs sales, stock, warehouse, returns, manufacturing and finance for a two-company manufacturing and distribution group, live in production. Two-way sync with Sage 50 accounts, a Kotlin Android app for warehouse handhelds, a customer ordering portal, and an in-app AI assistant (Claude) that answers questions across the business through 17 tools on its own MCP server.

The problem

An industrial pump distributor and its sister precision-engineering manufacturer ran on spreadsheets, paper and a desktop accounts package. Nobody could say with confidence what was on the shelf.

What I built

Web platform in Python/Flask with REST and WebSocket APIs: sales and purchase orders, stock, warehouse picking and despatch, returns and RMA, bills of materials and works orders, quotes, invoices, and a customer ordering portal.
Two-way accounts sync: a C#/.NET Windows service (~40k lines) keeps the platform and Sage 50 in step over an authenticated WebSocket.
Warehouse app: Kotlin/Jetpack Compose on Zebra RFID handhelds, with barcode and RFID label printing.
AI assistant in production: staff ask plain-English questions, from what is on the shelf to the profit on any customer over any period. Claude answers through 17 business tools served by the platform's own MCP server, builds custom reports on demand and helps train new users.
Release engineering: one container image rolled out to four Kubernetes environments (test and production for each company), with pre-flight gates, schema migration on every rollout and a post-deploy drift check.

Results

Accurate real-time stock for the first time in the companies' history, and fewer despatch errors.
Two businesses on one platform.
370 database migrations shipped with no data loss; about 3,300 automated tests.

Private system: a walkthrough or live demo is available on request.

AI Business Platform

Technology