Marketplace · United States and Canada
A Django marketplace where roasters buy in-stock green coffee from verified suppliers: multi-role accounts, a 7-state order workflow, warehouse-aware shipping and Stripe Connect payouts under separate US and Canadian configurations.
A spot marketplace for green coffee. Roasters, cafés and restaurants describe what they need, from origin and cup score to bag count and delivery state, and get matched with in-stock lots from verified importers and farmers across the US and Canada. No imports to arrange, no waiting; the coffee is already in a warehouse and ready to ship. Suppliers list lots, manage inventory and get paid through the platform.
The founders had a clear thesis and a market that still ran on emails and spreadsheets. They needed a backend that could take a first version to market quickly, but B2B coffee trade isn't a simple cart: the same company can buy and sell, one cart can draw from several warehouses, shipping rules differ by region and weight, and sellers in two countries have to be paid under different Stripe accounts and tax regimes. We led backend development from the MVP onward and have stayed on as the platform's engineering partner.
How the system is layered, from the people using it down to the data and the services it depends on.
The choices that made the difference, and why we made them.
Orders move through seven states from placed to fulfilled, refunded or cancelled. Allowed transitions are enforced in the serializers, so a client can't skip a state or reverse one by sending the wrong payload. When a single cart spans several warehouses, the order is split atomically into per-warehouse orders and earnings so each seller sees exactly their part.
US and Canadian sellers are paid through independent Stripe Connect configurations with their own webhook endpoints. An earnings ledger records every sale, clears the marketplace commission automatically and lets sellers trigger their own payouts, with the country-specific account resolved from the seller's profile rather than scattered through the code.
A dynamic shipping engine uses warehouse templates, weight tiers and region-based rules so grouped orders are priced correctly. Cart validation re-checks prices and stock at every step and adjusts inventory in real time, which stopped the oversold-lot problem before launch.
A DB-tracked scheduled task dispatcher sits over Celery, adding retries and status visibility that plain task scheduling lacks. A self-authored crash-alerting system emails admins full error reports on unhandled exceptions. For an early-stage marketplace it was the right trade: full visibility, no extra vendor.
Tell us what you're building. You'll get a straight answer on scope, timeline and cost, usually within one business day.