Savvy — AI-Powered Fashion Marketplace

homepage
second view of the homepage
cart page
chat / messag page
ai page
messaging section with live messaging
products page
seller product category page
seller chat / message page
seller orders page
seller ai page
seller dashboard
seller products page

Inspiration

Fashion shopping in Ghana and across Africa is fragmented — buyers scroll endlessly through social media DMs, sellers struggle to manage inventory, and there's no intelligent layer connecting the two. We wanted to build something that felt like a premium, modern marketplace but was built specifically for this market.

What it does

Savvy is an AI-powered fashion marketplace that connects buyers and sellers in real time.

For buyers:

Browse curated fashion products with smart search by keyword, location, or image
Get personalized AI fashion advice and outfit recommendations
Chat directly with sellers — a product card is automatically shared when a conversation starts
Add to cart, place orders, and track deliveries

For sellers:

A full seller dashboard with live inventory analytics, revenue charts, and product views
AI tools: listing generator, price suggester, inventory analysis, visibility tips, and market trends
Real-time inbox — buyers' enquiries arrive instantly with the product card attached
Manage products, categories, stock, and orders from one place

How we built it

Layer	Technology
Frontend	React, TypeScript, TailwindCSS, Recharts, shadcn/ui
Backend	Java Spring Boot, Spring Security (JWT), WebSocket (STOMP)
Database	PostgreSQL (Hibernate/JPA)
Media	Cloudinary (product images, profile pictures)
AI	Anthropic Claude API
Tunneling	ngrok (development)

The chat system uses WebSocket with STOMP — when a buyer starts a conversation, the backend automatically saves a structured product card as the first message and pushes it to the seller's inbox in real time.

The AI layer is powered by Claude and handles fashion advice, product listing generation, price suggestions, inventory analysis, and market trend insights.

Challenges we faced

Real-time chat architecture — designing the WebSocket flow so product cards appear instantly in the seller's inbox without polling
Multi-role authentication — buyers and sellers have separate auth flows, JWT principals, and permission layers within the same Spring Security config
AI response parsing — Claude returns structured JSON for inventory analysis; building a frontend renderer that gracefully handles both plain text and structured data responses
Image handling — managing Cloudinary upload flows for products with multiple images, display ordering, and efficient retrieval

What we learned

How to build a production-grade real-time messaging system with Spring WebSocket and STOMP
Prompt engineering for structured AI responses and how to parse them reliably on the frontend
Designing multi-tenant seller/buyer systems with clean separation of concerns

What's next

Mobile app (React Native)
Payment integration (Paystack/MTN Mobile Money)
Order tracking with live delivery updates
AI-powered size recommendations
Seller analytics with deeper sales trends

Built With

Submitted to

API Innovate 2026

Created by

I developed Savvy entirely as a solo full-stack project, handling the design, architecture, and implementation across all layers of the system.

I built the backend using Spring Boot, implementing JWT-based authentication for multi-role users (buyers and sellers), REST APIs, and a real-time messaging system using WebSocket (STOMP). I designed the chat flow so that when a buyer initiates a conversation, a structured product card is automatically generated and delivered instantly to the seller.

On the frontend, I developed the user interface using React, TypeScript, and TailwindCSS, creating a seamless experience for both buyers and sellers, including dashboards with analytics and real-time updates.

I integrated AI features using **ASI**, powering product listing generation, price suggestions, inventory insights, and personalized fashion recommendations, and built logic to handle both structured and unstructured AI responses.

I also handled media management with Cloudinary, database design with PostgreSQL, and overall system integration, ensuring the platform works as a cohesive, production-ready application.

This project reflects my ability to independently design and build a complex, scalable, real-world product from scratch.

Cyril-dot Ottih

Updates

Cyril-dot Ottih started this project — Mar 19, 2026 02:25 PM EDT

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.