02 / marketplace
Build trust into the transaction, not around it.
Status
Live deployment
Category
Commerce platform
Video walkthrough
Negus Gebeya — Marketplace in motion.
A buyer discovers or searches listings, reviews the seller and trust signals, saves or messages the seller, places an order through the payment flow, and follows the order state. A seller can create listings with media, manage orders, and use assisted listing tools.
A visual tour of the primary product flow
Context
Negus Gebeya is an online marketplace for Ethiopia, spanning electronics, vehicles, fashion, home goods, and more.
The problem
A marketplace has to make discovery, seller confidence, communication, and checkout work together. A listing page alone is not the product.
My contribution
I built the product surface and the backend boundary: listing creation and editing, search and filters, seller profiles, favorites, notifications, messaging, orders, payments, admin moderation, and the integration seams for AI and live selling.
Real workflow
A buyer discovers or searches listings, reviews the seller and trust signals, saves or messages the seller, places an order through the payment flow, and follows the order state. A seller can create listings with media, manage orders, and use assisted listing tools.
Architecture
The backend README describes a Bun + Hono + TypeScript API backed by MongoDB/Mongoose, with Cloudinary for media, JWT sessions, Zod validation, WebSockets for messaging, and an RTMP media server for live streams.
Integrations
Chapa checkout, Cloudinary media, WebSocket messaging, live-stream infrastructure, and AI-assisted listing workflows are treated as explicit service boundaries rather than hidden UI promises.
What works in the described product
- Auth with cookie or bearer JWT sessions
- Listings with multi-image and video uploads
- Seller profiles, trust scores, ratings, and verification workflow
- Favorites, notifications, direct real-time messaging, and reporting
- Chapa checkout, order tracking, seller earnings, and withdrawals
- AI listing assistant and live-selling route families
Verification and boundaries
The public frontend repository is delivered as a source archive and the public backend repository documents the route families, features, stack, and structure. The live URL is included so the product can be discussed as a workflow, not just a screenshot.
Next: For a client delivery, I would make the payment, media, and streaming environments explicit in the deployment runbook and run a staged buyer-to-seller transaction pass before production launch.