StopDown
Meesho logo
Launch · 4 June 2026Meeshosocial commerce platformmeesho.com ↗

Meesho launches AI system PRISM to drive 75% of orders via discovery-based feed

By · Editor ·

Bengaluru
  • Meesho uses an AI system called PRISM to recommend products to its 264 million transacting users each year
  • PRISM handles 75% of orders by predicting user intent through real-time behavior, not just text searches
  • The system processes 6 trillion daily inferences and uses 100 AI models trained on 400 trillion signals

What happened

  • Meesho launched PRISM, an AI system that drives 75% of its orders through a discovery-based feed.
  • PRISM uses real-time user behavior to predict intent, unlike traditional e-commerce search bars.
  • The system processes 6 trillion daily inferences and uses 100 AI ranking models trained on 400 trillion input signals.
  • During peak traffic, PRISM handles 100 million inferences per second.
  • Meesho built BharatMLStack, an in-house machine learning infrastructure to reduce data processing costs.
  • PRISM includes Trendpulse, an LLM that tracks regional purchase patterns and matches inventory to local demand.
  • The system supports 10 languages, focusing on voice-led navigation and regional dialects.
  • PRISM manages product recommendations, trend analysis, and seller distribution metrics.
  • It operates on data from 263 million monthly active users, 17 billion daily product views, and 2 billion user ratings and reviews.

Why it matters

  • Meesho’s PRISM system reflects a shift from keyword-based search to intent-driven discovery in e-commerce.
  • The system’s ability to handle 6 trillion daily inferences highlights the scale of Meesho’s user activity.
  • BharatMLStack was developed to cut costs compared to standard cloud solutions.
  • Trendpulse helps align inventory with local demand, improving relevance for users in different regions.

The business

  • It focuses on India’s growing internet user base, many of whom prefer voice commands and browsing over text search.
  • The platform processes 17 billion daily product views and 2 billion user ratings and reviews.
  • Meesho’s Net Merchandise Value (NMV) reached ₹11,371 crore in Q4 FY26, up 43% year-on-year.
  • The company reported a 66% reduction in losses, with contribution margins rebounding to 4.0% of NMV.

The founders

Debdoot Mukherjee

Chief Data Scientist
  • Stated that incoming internet users prefer browsing and voice commands over text search.

In numbers

₹11,371 croreNet Merchandise Value (NMV) · Q4 FY26
43%NMV growth · Q4 FY26
43%Orders growth · Q4 FY26
66%Loss reduction · Q4 FY26
4.0% of NMVContribution margin · Q4 FY26
263 millionMonthly active users · 2026
17 billionDaily product views · 2026
2 billionUser ratings and reviews · 2026

Up against

FlipkartAmazon IndiaSnapdeal

In their words

“Incoming internet users prefer browsing and voice commands over text search, requiring systems that interpret contextual behavioral data rather than explicit search queries.”Debdoot Mukherjee, Chief Data Scientist

Words to know

AI
Artificial Intelligence, a field of computer science that enables machines to perform tasks that typically require human intelligence.
NMV
Net Merchandise Value, a metric used in e-commerce to represent the total value of goods sold on a platform.
LLM
Large Language Model, a type of AI that processes and generates human-like text based on vast amounts of data.
BharatMLStack
An in-house machine learning infrastructure developed by Meesho to reduce data processing costs compared to standard cloud options.

The numbers around it

Commerce & Consumer Brands in June 2026: 24 rounds with a disclosed amount, $298 million in all; the biggest was FirstClub ($55M). Every sector this year →

Read it as a card in the app

Everything on Meesho → · All startup news on 4 June 2026 →

Is this Meesho? Get your “Featured on StopDown” badge →