PRISM, the Personalised Ranking & Intent Signal Module, continuously analyzes behavioural, transactional, and contextual signals to determine user intent in real time, running on more than 100 ranking models and processing over six trillion inferences daily.

During peak sale events, PRISM handles about 100 million inferences per second to enable real-time personalized discovery and proactive product recommendations.

The system operates at scale to connect sellers with users most likely to engage and purchase, delivering services in more than 10 languages, including Hindi, Bengali, Marathi, Tamil, Telugu, Kannada, Malayalam, Gujarati, Punjabi and Odia.

Meesho’s ecosystem supports 263 million monthly active users, 17 billion daily product views, 1,592 million ratings, and 505 million reviews, which bolster PRISM’s recommendation and discovery capabilities.

PRISM runs on BharatMLStack, Meesho’s in-house ML infrastructure, and includes Trendpulse, an LLM-powered discovery engine that identifies emerging regional demand trends.

PRISM currently supports features across recommendation systems, discovery, trend intelligence, catalogue understanding, and seller growth, aiming to make product discovery more natural and efficient for both consumers and sellers.

Meesho reports that over 75% of its orders now come from PRISM’s AI-powered discovery, signaling a shift from keyword searches to personalized discovery.