Meesho just tore down a major wall in India’s AI space. By throwing open the doors to BharatMLStack, its internal machine learning engine, the company isn’t just sharing code—it’s making a bold statement. For the first time, Indian startups and developers can peek under the hood of a battle-hardened, production-grade AI stack that’s been years in the making.
With critical components like the online feature store, control plane, orchestration tools, and SDKs now live on GitHub, Meesho is placing powerful tools in the hands of lean teams, scrappy coders, and ambitious data scientists. BharatMLStack was born out of necessity—real-time ML infrastructure that could perform at India-scale without setting fire to the balance sheet. And it delivered. In FY2025, it devoured 1.91 petabytes of data daily, pulled off 66.9 trillion feature retrievals, and enabled 3.12 trillion inferences, even thriving under the chaos of Meesho’s Mega Blockbuster Sale.
But what makes this move thrilling is Meesho’s intent. There’s no price tag, no slick monetization strategy. Just an open invite to build, break, and better the system together. The IPO-track e-commerce giant wants BharatMLStack to spark grassroots-level AI evolution in India, giving even the smallest startups access to high-performance ML tech.
This isn’t a press release moment. It’s a ripple that could spark a wave—one where Indian innovations aren’t throttled by infrastructure.
1. The Genesis of BharatMLStack
1.1 Purpose-Built AI Infrastructure
Meesho didn’t stumble into building BharatMLStack—it was born from frustration. Their earlier ML systems were a Frankenstein’s monster of pricey proprietary tools and clunky open-source ones. Nothing quite fit. So they built their own. The result? A central platform that slashed inefficiencies, handled complexity, and didn’t crumble under pressure. It’s real-time, rugged, and ruthlessly optimized.
1.2 Designed for Indian Conditions
India’s digital landscape is messy—internet speeds fluctuate, devices are low-spec, and people type search queries in Hinglish. BharatMLStack is tuned to all that chaos. It doesn’t just survive these quirks—it thrives in them. It keeps predictions sharp, latency low, and costs lean. And that’s what makes it such a game-changer for Indian startups.
2. Meesho’s Operating Model and Vision
2.1 Business Model Overview
Meesho’s core idea is democratizing e-commerce. Sellers don’t pay commissions. Instead, the company earns through logistics partnerships and value-added services. It’s like giving the tools to sell online to every kirana store, homemaker, and small-town hustler. The ML infrastructure it’s now open-sourcing was built to serve this very diversity at scale.
2.2 Revenue and Funding
While FY25 numbers are under wraps, Meesho has hit profitability in multiple quarters before. The big money—over $1 billion—came from power players like SoftBank, Facebook, and Prosus Ventures. Add to that a reverse flip back to India and whispers of an IPO, and it’s clear: Meesho isn’t playing small.
3. What Problems Does BharatMLStack Solve?
3.1 Infrastructure Cost and Scalability
Most startups can’t afford to build an ML pipeline from scratch. BharatMLStack takes that pain away. It’s plug-and-play, optimized for speed, and helps engineers deploy models fast—without a PhD in infrastructure. Less cost. More firepower. No silos.
3.2 Real-Time ML Use Cases
Think fraud detection that works before damage is done. Recommendations that feel eerily personal. Automated tagging that just gets it. Meesho built these use cases for itself—now others get to leverage them. For scrappy founders, this is like getting a Ferrari engine for a startup scooter.
4. Technical Performance Metrics
4.1 Data Processing at Scale
Let’s talk scale. BharatMLStack chews through nearly 2 petabytes of data daily. That’s not a typo. 66.9 trillion features get retrieved, and 3.12 trillion inferences fire off—every day. These aren’t vanity metrics. They’re proof that the stack holds up when the heat is on.
4.2 Proven at Peak Load
During the Mega Blockbuster Sale, traffic spiked—and BharatMLStack didn’t blink. Not only did it stay up, it helped drive more clicks, better conversions, and higher order volumes. When the rubber hit the road, it delivered.
5. Strategic Importance of Open-Sourcing
5.1 A Non-Monetized, Community-Driven Approach
This isn’t about cash. Meesho isn’t trying to turn BharatMLStack into a product. It wants startups to use it, challenge it, improve it. Even if just a dozen companies plug it in and build something magical, the mission’s a win. This is about shared progress, not quarterly returns.
5.2 Democratizing AI in India
India’s AI scene is bursting with talent—but not everyone has access to top-shelf infrastructure. Meesho’s move levels the playing field. Whether you’re bootstrapping in Pune or scaling in Bengaluru, you now have access to tools that were once locked inside a unicorn.
6. Meesho’s Broader Impact and Competitive Landscape
6.1 Empowering AI Startups
India wants to be an AI powerhouse. But to get there, startups need more than ambition—they need tools. BharatMLStack delivers those tools, minus the enterprise price tag. It’s fuel for the next generation of AI-first founders.
6.2 Competitors and Differentiation
Sure, Amazon and Flipkart are rivals in e-commerce. But BharatMLStack also takes a swing at the Googles and Microsofts of the world—specifically their ML platforms. The difference? Meesho’s stack isn’t generic. It’s built for India, by people who know its quirks.
7. Meesho’s Founders and Company Journey
7.1 Founding Story
Vidit Aatrey and Sanjeev Barnwal weren’t chasing after the next big, flashy startup dream. They saw firsthand how brutally tough it was for small businesses to actually make it online and decided to step up and do something about it. That mission? It’s never wavered. In fact, it’s only gotten more daring.
7.2 Building with Purpose
Every product Meesho rolls out has India baked into its DNA. BharatMLStack isn’t an exception—it’s a culmination. From reselling app to e-commerce engine to AI enabler, Meesho’s journey is proof that Indian startups can think big, build deep, and still stay rooted.
8. Industry Trends and Future Outlook
8.1 AI and ML Adoption in Startups
India has over 3,000 AI startups, says NASSCOM. But most of them are stuck wrestling with infra problems. If BharatMLStack gains traction, we could see a new wave—startups that are lean, fast, and ML-native right from Day 1.
8.2 Future Releases and Open Roadmap
More modules are on the way—model serving, registration, workflows. Meesho’s taking a phased approach so the ecosystem can grow with the stack. No rush. Just momentum.
Learning for Startups and Entrepreneurs
Speed is great. So is scale. But what’s rarer is infrastructure that does both without breaking the bank—or the team. BharatMLStack shows what’s possible when you build smart and share openly. For founders, the takeaway is clear: if you’ve built something great, don’t hoard it. Share it. Shape the future with it.
About Foundlanes
At foundlanes.com, we spotlight stories that shake things up in India’s startup space. And “Meesho open-sources in-house ML platform, impacting Indian startups” is one of those moments. It’s about more than AI—it’s about access, openness, and community-first innovation.
If you’re building something bold, we’re here to tell your story. From tech disruptions to grassroots revolutions, The Startups News keeps you connected to India’s startup heartbeat.