Full talk · edited transcript

Building Multimodal Product Embeddings to Improve Matching at Yandex Market

Boris Ivanchikov7 Nov 2025

Boris Ivanchikov, ML Technologies Development Service (ML R&D), Yandex.

The ML R&D team helps Yandex services (Market, Eda, Lavka, Taxi) with ML projects. The Market matching team came to them with a request: build multimodal product embeddings. The talk covers what matching is, why a multimodal encoder is needed, how it is built, how it was trained, and what was achieved.

Continue reading — it's free

Google or a one-time email code. No password.