CodeXhange
ActiveGamified promo code discovery.
Search engine · E-Commerce · My role: Architected & Built
What it is
CodeXhange is a high-throughput promo code search engine and continuous merchant coupon harvester designed to eliminate broken discounts, expired vouchers, and affiliate cloaking across 10,000+ brands.
Why it exists
Most coupon platforms on the internet are filled with affiliate junk, deceptive redirects, and dead codes. Shoppers waste minutes copying broken strings at checkout. CodeXhange was created to bring transparency to savings with instant search, automated voucher verification workers, and community confidence scoring.
What I did
- Architected and engineered the end-to-end application stack using Next.js, Node.js, Express, MongoDB, and Redis
- Engineered continuous background scrapers and verification crawlers validating discount codes against live merchant checkout endpoints
- Implemented high-concurrency compound indexing in MongoDB for sub-50ms fuzzy search across millions of discount entries
- Built a real-time reputation scoring algorithm that automatically downgrades and purges non-working codes based on user success reporting
- Implemented Redis caching layers with intelligent TTL invalidation for trending brands and seasonal shopping spikes
Technical decisions that mattered
Compound text & merchant indexes with Redis read-through caching
Queries hit a multi-field compound index spanning merchant slug, active status, expiration timestamp, and verification score. High-traffic brand pages are cached in Redis with a sliding TTL, ensuring 99th-percentile response times remain under 40ms.
Decoupled harvesting pipelines from query serving
Crawling and merchant DOM inspection are handled by isolated background queue workers. The user-facing search engine never blocks on slow external merchant servers or network timeouts during coupon validation.
Defensive normalization for merchant voucher strings
Input promo codes pass through strict canonicalization (whitespace trimming, case normalization, zero-width character stripping, and pattern matching against merchant coupon formats) before hitting the index.
What I learned
The web is noisy and web harvesting without defensive schemas quickly degenerates into garbage data. Rigorous validation rules and an automated reputation scoring loop are what keep search results accurate.
Status
Active — continuous indexing