Meesho open-sources BharatMLStack to boost AI innovation

Meesho has open-sourced key elements of its internal machine learning platform, BharatMLStack.

The components, including an online feature store, control plane, orchestration UI, and SDKs, are now available on GitHub.

Built over two years, BharatMLStack supported 66.9 trillion feature retrievals and 3.12 trillion real-time inferences during FY25.

The online feature store helps power applications like fraud detection and personalized recommendations using real-time and precomputed features.

Meesho said that the open-sourced platform is designed to simplify and accelerate machine learning deployment for developers.

The company also plans to release future components of BharatMLStack as open-source, inviting contributions from the developer community.

🔗 Source: YourStory

🧠 Food for thought

1️⃣ Open-source ML is redefining competitive dynamics in tech

Meesho’s decision to open-source BharatMLStack follows a significant industry shift in how companies approach proprietary technology.

Open-source ML platforms have evolved from academic origins into corporate strategy tools, with NumPy (1995) and scikit-learn (2007) paving the way for today’s collaborative development environment 1.

This approach represents a strategic evolution from traditional competitive dynamics where companies would keep core technology proprietary to gain market advantage.

By open-sourcing infrastructure that powered 66.9 trillion feature retrievals, Meesho gains community contributions while positioning itself as an innovation leader, a pattern previously seen when Google released TensorFlow (2015) and Facebook released PyTorch (2016) 2.

Companies embracing open-source ML typically focus monetization on cloud infrastructure rather than software licensing, reducing costs while accelerating development through community scrutiny 3.

For IPO-bound Meesho, this move potentially enhances their technical reputation while solving the challenging “build vs. buy” dilemma that many companies face with ML infrastructure.