Pinterest uses an AI system called OmniSearchSage to power its search results and shopping recommendations. Unlike earlier systems that treated text searches and image searches as separate problems, OmniSearchSage places search queries, Pin images, and product listings into a single shared mathematical space so that a search for "minimalist bedroom" can surface both text-matching articles and visually relevant room photos at the same time. The system serves approximately 300,000 search requests per second..

OmniSearchSage is a multi-task, multi-entity embedding model: it learns a single unified representation space for three types of items — search queries (text), Pins (images + text), and products (catalog items) — so that similar items end up close together regardless of their format. The model was described at The Web Conference (WWW '24) in April 2024. Pinterest reports improvements of more than 8% in search relevance, more than 7% in user engagement, and more than 5% in ad click-through rate following deployment. Additionally, Pinterest uses a large language model distillation pipeline — where a large model (LLaMA-3 8B) teaches a smaller, faster model to score relevance — to further improve search quality with less computational cost.