Adventures of integrating multimodal search in e-commerce
16:00 - 16:30, 27th of May (Tuesday) 2025 / DEV AI & DATA STAGE
Vector search in 2025 is a table-stakes technology.
However, properly integrating multi-modality into an existing e-commerce search application remains a big challenge.
And with scale, the challenge only gets harder.
In this talk, I want to share my experience of integrating multi-modal search capabilities at Vinted.
I'll list the trade-offs of combining nearest neighbour search with support for filtering. Also, I'll dive deep into the toughest business requirements and how they were implemented. Furthermore, we'll touch on Vespa architecture and how it allows performant semantic search at billion-scale. It's going to be a wild ride because we have to cover a lot.
LEVEL:
Basic
Advanced
Expert
TRACK:
AI/ML
Data
Software Architecture
TOPICS:
AI
Bigdata
ITarchitecture
ML/DL
