Meesho Open-Sources BharatMLStack to Power India’s AI Future

In a landmark move to democratize access to high-performance AI infrastructure, meesho has open-sourced key components of its in-house machine learning platform, BharatMLStack, on GitHub. This makes the IPO-bound e-commerce company one of the first major Indian players to contribute to the country’s AI ecosystem at such a significant scale.

The release includes a production-grade feature store, control plane, orchestration UI, and SDKs—modules designed to power real-time, scalable machine learning applications. Developed over the past two to three years, BharatMLStack is already a core part of Meesho’s infrastructure, handling a staggering 1.91 petabytes of data per day, 66.9 trillion feature retrievals, and 3.12 trillion real-time inferences during peak loads in FY25.

“We put BharatMLStack to the test during high-traffic events like our Mega Blockbuster Sale in March 2025, where it delivered at scale,” said Sanjeev Kumar, Meesho’s Founder and CTO. “By open-sourcing it, we’re sharing a high-scale AI stack tailored for real-time use cases and Indian businesses.”

The platform supports critical use cases like fraud detection, personalized search, product ranking, and automated tagging, with built-in sensitivity to Indian user behaviors—like transliterated searches and usage on low-resource devices.

Debdoot Mukherjee, Meesho’s Chief Data Scientist, emphasized the limitations of expensive or rigid proprietary tools: “With proprietary platforms, it’s like driving a sports car when all you need is a compact vehicle—and you can’t even look under the hood. Open source gives you transparency and control.”

The initiative also aims to empower early-stage startups, independent ML engineers, and resource-constrained teams. Meesho isn’t looking to monetize the release, instead seeking feedback, collaboration, and community-driven innovation.

The current release starts with the Online Feature Store, essential for low-latency tasks like recommendations and fraud detection. More components—including model serving and workflow tools—will be released in phases.

Meesho hopes that even if “a dozen companies” adopt the stack in their workflows, the project will be a success.

With BharatMLStack, Meesho is not only building technology for scale—it’s building an ecosystem for India’s AI future.