Case Study: Building a Zero-Friction B2B Network for Connected Warehouses
Architected a multi-tenant platform that lets warehouses sell to dealerships directly through their existing ERPs, featuring zero-friction QuickBooks onboarding and LLM-powered order predictions.
Project Information
Role: Technical Architect • AI Product Engineer • Backend Engineer
Industry: Automotive Parts • B2B Commerce • ERP Integration • AI
The Problem
Warehouses struggled to scale sales because selling to external dealerships required manual entry across conflicting software. Dealerships had to browse dozens of isolated portals to find stock, while warehouse admins lacked any predictive tools to forecast costs or order demand.
Technical Solutions & Direct Results
Zero-Friction QuickBooks Onboarding: I built automated background data pipelines that pull legacy products, past orders, and customer profiles instantly upon connection, eliminating manual cleanup for new warehouses.
Multi-Tenant ERP Architecture: I designed an isolated database schema that safely hosts multiple ERP networks simultaneously, enabling the platform to scale to new accounting software without risking data leaks.
LLM Order Forecasting & Cost Tracking: I developed an AI data pipeline that analyzes historical purchasing habits to predict future dealership stock needs, populating an admin dashboard with automated, itemized cost margins.
The Single-Portal Experience: I engineered a real-time synchronization engine for products, prices, discounts, orders, and invoices, allowing dealerships to buy from a wide network of warehouses within a single portal.
The Bottom Line
For Warehouses: Sell and ship inventory globally without ever leaving your primary, day-to-day ERP software.
For Dealerships: Source from an endless aisle of warehouse inventories through a single, unified screen.
Technical Stack
FastAPI
PostgresSQL
OpenAI
Cursor, ClaudeAI & Codex
QuickBook