NuvveTech Marketplace

01 · Context

What the platform needed.

NuvveTech runs a live e-commerce marketplace connecting vendors and customers across multiple regions. A marketplace at that stage needs far more than product listings: payments, fulfillment, search, supplier integrations and performance all have to work reliably together, and a failure in any one of them shows up directly to real customers.

02 · Engineering challenge

What made the work meaningful.

Search needed to go beyond exact keyword matches so customers could find products by what they actually looked like, supplier data had to flow into the platform automatically instead of by hand, and the backend had to stay fast as the catalog and order volume grew.

03 · Contribution

What Andrew worked on.

Andrew worked as a contract backend engineer on NuvveTech's production systems, focused on payments, search, supplier integrations and performance. His work included the Stripe payment processing flow, an automated order-tracking pipeline, an AI-powered image search feature, and a broader pass on backend performance.

04 · Engineering approach

How it was built.

Payments and order tracking were built around Stripe's webhook events, so order status stayed accurate without manual reconciliation. For search, product images were indexed with a FAISS vector index built on CLIP embeddings, so customers could find visually similar products in addition to keyword search, backed by PostgreSQL full-text search for standard queries. Supplier integrations pulled catalog and inventory data automatically instead of relying on manual updates, and Redis caching plus Celery background jobs took repeated, expensive work off the request path.

05 · Key capabilities delivered

What shipped.

• Stripe payment processing with webhook-driven order updates
• Automated order-tracking pipeline
• AI-powered visual product search using FAISS and CLIP
• PostgreSQL full-text search for keyword queries
• Third-party supplier catalog and inventory integrations
• Redis caching and Celery background processing
• Backend performance work: N+1 query elimination, query optimization and response caching

07 · Outcome

What it delivered.

• Production payment processing integrated directly into the order flow.
• Customers could search visually, not just by keyword, improving product discovery.
• Supplier data stayed current without manual catalog updates.
• Backend response times improved after eliminating N+1 queries and adding response caching.