Meesho open-sources key components of BharatMLStack for wider AI adoption

E-commerce platform Meesho has open-sourced key components of its internal machine learning platform, BharatMLStack, to support AI infrastructure access for Indian startups and developers.

The release, now live on GitHub, includes Meesho’s feature store, orchestration interface, control plane, and software development kits. Built over two years, BharatMLStack supports the company’s daily data processing needs, which total around 1.91 petabytes, and enables real-time applications.

In FY25, Meesho’s ML systems recorded 66.9 trillion feature retrievals and delivered 3.12 trillion real-time predictions. These systems supported operations during events such as the Mega Blockbuster Sale held in March 2025.

By sharing its feature store, Meesho aims to help data teams build AI systems for real-time use cases such as fraud detection and recommendation engines. The feature store provides consistent access to precomputed data signals and reduces the time required to deploy models.

Meesho plans to release additional parts of BharatMLStack in future updates. The platform is now open to developer contributions via GitHub.

The company also completed its reverse flip to India after receiving final approval from the National Company Law Tribunal last month.