AI E-Commerce Platform
A complete store with AI visual recommendations: upload an image and the AI identifies the event, trip or location, then suggests matching products. Secure payments with Razorpay and Stripe.
An end-to-end e-commerce platform where shoppers don't have to know what to search for. Upload a photo — a beach trip, a wedding, a hiking spot — and the AI works out the occasion or place and recommends products that fit it.
The problem
Most online stores rely on keyword search, which only works when you already know what you want. People often shop for an occasion instead: "what do I need for a Goa trip?" or "what should I wear to this event?" I wanted to turn a single image into a set of relevant products.
What I built
- AI visual recommendations: users upload an image; the AI identifies the event, trip or location in it and surfaces matching products from the store's catalogue
- Full shopping flow: product browsing, search, cart and checkout
- Secure payments: integrated both Razorpay and Stripe, so the store works for Indian and international customers
- Admin panel: manage products, inventory and order status from one place
- Authentication: user accounts with order history
How the AI works
- The user uploads an image from the shopping page.
- The React frontend sends it to the Node.js/Express backend.
- The backend passes it to a Python AI service, which analyses the image and returns the detected context (for example, "beach", "wedding" or "mountain trek").
- The backend maps that context to product categories and tags in MongoDB.
- Matching products are returned and shown to the user as recommendations.
Tech stack
- Frontend: React, Tailwind CSS
- Backend: Node.js, Express.js
- Database: MongoDB
- AI: Python
- Payments: Razorpay, Stripe
Challenges
- Turning AI output into products: image analysis returns loose descriptions, so I built a mapping layer that converts detected contexts into the store's own categories and tags.
- Supporting two payment gateways: Razorpay and Stripe handle orders and confirmations differently, so I wrapped both behind one checkout flow.
- Keeping it fast: image analysis takes time, so the UI shows clear loading states and the rest of the store stays usable while recommendations load.
What I learned
How to connect a Python AI service to a MERN application, design a recommendation flow around fuzzy AI output, and build a payment system that handles more than one provider cleanly.