Evolving Search in Online Marketplaces: Practical Applications of AI
summary
Velebit AI describes concrete systems: image taggers (color and pattern APIs), fine‑grained categorization for marketplaces with thousands of categories, and duplicate detection tuned per client; a 12‑person team uses CNNs, transformers, CLIP and OpenAI embeddings with vector DBs, BM25 and sparse embeddings in a hybrid pipeline. Iterative LLM prompt refinement powers interactive text→image search, cross‑encoder rerankers improve top‑K, and earlier 2016 prototypes predate commercial offerings.
AI-written summary · about
chapters (13)
- 0:00Introduction to AI search
- 1:54About Valit AI
- 3:00Challenges in marketplace search
- 4:04Preparing products for search
- 8:01Search without product tags
- 10:07Interactive and real-world search
- 13:39Building modern search solutions
- 15:27Traditional keyword search
- 18:20Dense and sparse vector search
- 21:54Combining search results
- 25:05Reranking search results
- 27:11Key takeaways and trends
- 29:42Q&A and closing
related talks
31mAI Search for Marketplaces: Experiences & Tech
29mDeveloping Visual AI Solutions for Online Marketplace
16mSolve for AI
13mThe State of AI Powered Search and Retrieval
33mMachine Learning in Visual Search
22mHarnessing image and text data for product matching
22mMachine Learning: The Engine Behind Modern E-commerce Search
39mAI-Powered Search in Modern E-Commerce