← Back to work
Case study · Product & AI search

najdiavto.com

A production car marketplace with real dealer inventory — high-cardinality faceted search plus a live natural-language AI search feature. The ML service behind it has its own case study.

200+car brands
212municipalities
AINL search live
The product

A car marketplace that has to actually work

najdiavto.com is a live marketplace with real dealer inventory, not a static demo. Buyers browse and search thousands of cars across a fragmented regional market, so the product has to balance powerful filters with a search experience anyone can use. AI search is one layer on top of a solid, production search and listings stack.

What shipped

Three things that make it a real product

Search

High-cardinality faceted search

Real dealer inventory with thousands of discrete attribute values (200+ brands, 212 municipalities) requires faceted filtering that stays fast and useful.

  • Filters for make, model, fuel, transmission, body type, color, doors, and Euro norm
  • Deep-linking and shareable filter states
  • Server-rendered listings that stay fast at scale
AI search

Natural-language listing search

Buyers can describe a car in plain language and get matching listings, not just a keyword match — turning a catalog into a conversational tool.

  • RAG-based retrieval over the live inventory
  • Shipped as a real production feature, not a demo
  • Complements the structured filters for power users
Regional coverage

Built for a fragmented market

A Slovenian car market split across many municipalities needs local context — coverage, pricing signals, and listing quality all vary by region.

  • Inventory organized across 212 municipalities
  • Live dealer feed keeps listings current
  • Designed for real buyers, sellers, and dealers
Stack

Tools used

ReactNode.jsMongoDBVercelRAG