SnobOutlet

Overview

SnobOutlet is a premium marketplace connecting luxury brand sellers with discerning buyers worldwide. The platform features real-time bidding, semantic search powered by vector embeddings, and a microservices architecture designed for high availability.

Technical Challenge

Building a real-time bidding system that could handle thousands of concurrent auctions while maintaining data consistency and providing instant updates to all participants. The system needed to support:

• Sub-second bid placement and confirmation
• Real-time price updates across all connected clients
• Atomic bid validation to prevent race conditions
• Horizontal scaling during peak auction periods

Approach

We implemented a two-layered microservices architecture separating the API gateway from business logic services. Key technical decisions included:

• Event-driven communication using RabbitMQ for auction state changes, ensuring all services stay synchronized
• Milvus vector database for semantic product search using image and text embeddings, enabling "find similar items" functionality
• Redis caching layer for hot auction data and user sessions, reducing database load during high-traffic periods
• Optimistic locking with MongoDB transactions for bid consistency, preventing double-bids and ensuring fair auctions

Architecture Highlights

The backend follows a clean separation of concerns:

• API Gateway - Rate limiting, authentication, request routing
• Auction Service - Bid processing, timer management, winner determination
• Search Service - Vector similarity search, full-text search, filtering
• Notification Service - WebSocket connections, push notifications, email alerts

Impact

Successfully launched on iOS App Store with real-time auction functionality supporting hundreds of concurrent bidders per auction. The semantic search feature increased product discovery by 40% compared to traditional keyword search.