Meesho’s Big AI Move: What is BharatMLStack Power?

In a significant and rather rare move for a high-growth Indian tech company, Meesho—best known for disrupting social commerce and preparing for a much-anticipated IPO—has turned enabler for the broader startup ecosystem. By open-sourcing core components of its proprietary machine learning platform, BharatMLStack, Meesho is not just flaunting its tech muscle, but inviting India’s growing AI community to build, scale, and innovate—without starting from scratch or breaking the bank.

For Indian startups, many of whom are building for a uniquely diverse and complex user base, this move could be a game-changer.

Opening Up AI: Why This Matters

At a time when AI and ML infrastructure is mostly gated behind expensive tools or enterprise-grade solutions, Meesho’s decision to open up its in-house tech stack is being seen as an act of democratization. The company has made available on GitHub the first batch of tools from BharatMLStack—including a production-grade feature store, a control plane, orchestration UI, and SDKs—making it one of the first Indian e-commerce giants to contribute at this level to the AI community.

This is not just about open-sourcing code. It’s about giving early-stage startups and independent developers access to infrastructure that has been battle-tested in real-world, large-scale applications.

What Exactly Is BharatMLStack?

Built over the past two to three years, BharatMLStack is Meesho’s internal machine learning platform designed to power real-time AI use cases—ranging from personalized search and fraud detection to product tagging and dynamic recommendations.

Here’s a glimpse into the scale it operates at:

• 1.91 petabytes of data processed per day

• 66.9 trillion feature retrievals

• 3.12 trillion real-time inferences during peak traffic events

If those numbers sound mind-boggling, that’s because they are. And the most compelling part? This infrastructure is now being offered, in parts, for free.

A Real-World Stress Test: March 2025 Sale

The credibility of any tech stack lies in how it performs under pressure. Meesho put BharatMLStack through fire during its March 2025 Mega Blockbuster Sale, one of the highest traffic periods of the year. Not only did the system hold up, but it also improved key business metrics like user engagement, conversions, and order volumes.

“Great technology should scale impact, not just infrastructure,” said Sanjeev Kumar, Founder & CTO of Meesho.

By open-sourcing it now, Meesho aims to pass on the impact potential to the wider community—especially those building real-time AI systems for Indian users.

Built for Bharat: Why Indian Startups Should Care

India’s internet users are unique—think low-cost smartphones, mixed-language queries, inconsistent data networks, and exploratory buying behavior. BharatMLStack has been designed keeping these nuances in mind. From transliterated search support to fuzzy product discovery and modular fraud detection systems, the platform addresses the very problems most Indian startups are trying to solve.

“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,” said Debdoot Mukherjee, Chief Data Scientist at Meesho.“Open source gives you transparency and control.”

What’s in the Open-Source Kit?

The first component being released is the Online Feature Store, which acts as the fuel tank for any ML engine—supplying precomputed features to ML models in real time. This helps reduce latency, avoid data silos, and maintain consistency across ML-driven functions like fraud detection, recommendation engines, and more.

And that’s just the start. Meesho has also promised to roll out more components over time, including:

• Model serving infrastructure

• Model registration systems

• Workflow authoring tools

These will be open-sourced in a phased manner, based on community feedback and adoption.

No Monetization Plans—Just Collaboration

In an era where data is gold and AI tools are often behind paywalls, Meesho’s move to open-source its AI stack—without any monetization plans—is bold and refreshing.

Instead of profits, Meesho is looking for participation. The goal is simple: get feedback, drive collaboration, and help a dozen or more Indian companies integrate BharatMLStack into their systems. That alone, Meesho believes, would signal success.

For early-stage startups that don’t have full-fledged tech or data science teams, the plug-and-play nature of BharatMLStack can be a huge leap forward. And for more mature companies looking to scale their AI efforts cost-effectively, this could eliminate the need to reinvent the wheel.

Open-sourcing BharatMLStack signals a shift in mindset—from building in silos to building in the open. For Indian startups, especially those navigating real-time machine learning challenges, this is more than just code—it’s a shortcut to scalability.

It’s also a wake-up call to other big tech companies: that contributing to the developer ecosystem is not just generous, but strategic.

As AI becomes a defining layer in every business, the need for shared tools, transparent systems, and community-driven innovation will only grow. Meesho’s move to open-source BharatMLStack is not just a technical decision—it’s a cultural one. It’s about saying, “Let’s build this ecosystem together.”

For the Indian startup ecosystem, this could be the nudge it needs to accelerate AI adoption—faster, cheaper, and more meaningfully.