Meesho has taken the first step in open-sourcing its internal AI framework, BharatMLStack, starting with a robust Online Feature Store now available on GitHub. Developed over three years, this production-ready store is engineered to manage real-time, cost-effective ML operations at scale. Meesho plans to expand forthcoming releases to include modules for model serving and workflow orchestration.
Built specifically for India’s demanding e-commerce landscape, BharatMLStack can handle millions of requests per second, process 1.91 petabytes of data daily, and support trillions of real-time inferences. It is the backbone powering a range of machine learning applications—recommendations, fraud detection, automated tagging, personalized search, and more—maintaining resilience even during high-traffic events like large sales. As Sanjeev Barnwal, Founder and CTO of Meesho, put it:
“By open-sourcing it, we’re sharing a high-scale, AI stack with the broader tech community—purpose-built for real-time use cases and tailored for Indian businesses.”
Unlike traditional enterprise ML architectures that depend on costly hardware, BharatMLStack prioritizes efficiency, flexibility, and openness, making it well-suited to India’s budget-sensitive tech ecosystem. Its design also caters to local usage patterns, including features such as transliterated search and fuzzy discovery, which are essential for multilingual and vernacular user behavior.
With this release, Meesho aims to help startups and mid-sized businesses overcome common machine learning hurdles. The plug-and-play infrastructure is intended to reduce complexity, lower costs, and accelerate innovation, all without tying users into proprietary systems. The shared stack provides practical, scalable solutions for companies grappling with real-time ML demands but lacking in-house engineering resources at massive scale.
This initiative not only strengthens Meesho’s position as a leader in AI-driven e-commerce innovation but also contributes to democratizing advanced ML technologies across India’s growing startup ecosystem. As BharatMLStack evolves to include serving and orchestration, the open-source community can look forward to a potent, end-to-end AI platform—designed locally, built for scale, and now available to everyone.
