Co-Founder & CEO, @razorpay
Bengaluru, India
Joined September 2009
- Tweets2K
- Following401
- Followers52.4K
- Likes6.2K
Pinned Tweet
4 billion payments. 3 trillion data points. One model trained on all of it.
Meet @Razorpay Vulcan - India’s first transformer-based AI foundation model for payments.
Built, trained and hosted in India, in partnership with @nvidia and @awscloud.
Until now, every payments problem was solved separately: routing, fraud, risk, personalisation and more.
We asked: What if one model could understand how money moves?
And like LLMs are trained on text to understand language, Vulcan is trained on payments to understand how money moves.
Already running in beta across 51,000+ businesses, Vulcan is delivering:
- 8–10% improvement in payment success rates - the ultimate measure of whether a payment simply works.
- 8x more international card fraud detected.
- 5x more fraudulent or disputed transactions identified.
- 1–2 lakh more purchases completed every month through better checkout personalisation.
And we’re just getting started. The best part? Every payment Vulcan sees makes the next one smarter.
We’ve spent years building the infrastructure that moves money for India. Now, we’re building the intelligence that understands and improves it. 🔥
It’s wild how many people still underestimate AI’s impact beyond coding. Reminds me of this 1995 Bill Gates interview.
Letterman got the laughs, Gates built the future.
If your security team isn’t having sleepless nights after the last 24 hours, you should be worried.
Cybersecurity is entering a period of absolute chaos. And this might be the new normal for a while before things stabilise.
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source!
Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable.
🔗 github.com/deepseek-ai/DeepI…
1/4
Growing up, I remember that most Indians couldn’t bake at home. Ovens were largely limited to commercial bakeries.
Today, almost anyone can own an oven. Yet most people still buy their cakes from bakeries.
Because being able to bake, wanting to bake, and being good at baking are three very different things.
HTML might be the new Markdown, as @trq212 pointed out early this year. But collaborating on HTML is still painful.
At @Razorpay, as agents started generating interactive docs and decks, we needed a way for humans and agents to collaborate on them in real time. So one engineer built SlashDocs. The rest of the team (and our AI engineer Slash) kept making it better.
Today, it has replaces google docs for 50%+ of all new docs at Razorpay, changing how humans and AI agents collaborate.
And now we're open-sourcing it!
We've spent years building tools for humans to collaborate. Now we need tools for humans and AI agents to collaborate.
AI agents are changing how we work and collaborate, but documents are still stuck in the pre-AI era.
Agents can now create beautiful, interactive decks, dashboards, and reports in HTML. But the moment you want to collaborate on them, everything breaks down.
Google Docs wasn’t built for rich, interactive HTML or AI agents. Feedback gets scattered across Slack and email. Every edit means going back to the agent, generating a new file, and sharing another link. No real-time collaboration. No single source of truth.
We built SlashDocs to fix this.
Think Google Docs, but built for AI agents and the documents they create. Host HTML and Markdown in one central hub. Share documents, manage permissions, and comment directly on the content, just like Google Docs.
Agents can read feedback and update the same document in real time. No duplicate files, no endless back-and-forth, and everyone stays in sync.
What started as a small internal experiment in one team has now scaled across @Razorpay. 28,000+ documents created. Over 50% of all documents generated at Razorpay now live on
SlashDocs.
And today, we’re opening it up to everyone. Self-hosted. Agent-agnostic. Fully open source.
AI agents have changed how we create documents. It’s time we changed how we collaborate on them.
Try it: razorpay.com/slash-docs/
Nikhil is right. But businesses respond to demand.
I’d love to see a massive wave of Indian youth experimenting with personal AI agents.
They may not be very reliable or even useful today. But that’s not the point. Play with them just because they’re fun. Push them to their limits. Make them do things nobody thought possible.
India has the world’s largest youth population. And that’s a huge advantage only if it’s curious.
India needs to get agent-ready
because agents remove the complexity of using the web and apps for millions of our people: both language and UX barriers.
Why people have to learn an interface, and different interfaces for each app? Most of the web is in English, a language not native to most of our population. Agents solve for this friction. In my article in todays Economic times, I offer six initial things to be addressed:
1. Companies need to make their interface agent friendly: Flipkart, BigBasket, other merchants block agents. We need them to open up APIs/MCPs, and offer easy connections with agents, to expose resources, tools and structured inputs to AI applications.
2. Ecommerce catalogs created for humans need an upgrade: A human can look at a product page and infer quality and reliability. An agent needs exact data and easier discovery. A minor change in a page layout or an unexpected popup can interrupt the entire process. Popups meant for cross-selling and upselling to humans are wasted on agents. ONDC, using the Beckn protocol, has machine-readable catalogs and structured orders, and we should leverage that.
3. Our government websites need an upgrade: they are famously torturous for humans to navigate, leave alone poor agents. Government interfaces can expose authorised, intelligible and interoperable ways for an agent to act on a citizen’s behalf: departments can publish safe-action APIs.
4. We need more intuitive mechanisms for agents to make payments. I shouldn’t have to give my agent my credit card. While India has UPI AutoPay and e-mandates. We need NPCI to launch its Universal Agent Protocol. Agentic payments will spur agentic commerce.
5.Five, we need a safe way to delegate authorisation when agents enter regulated environments to transact.
6. We need to solve for liability: Agents can make mistakes and their behaviour is rarely deterministic. We need to manage risk, but if we stack on liability, people will stop using agents. Agent-insurance and easier reversibility of actions might be solutions.
Lets make India Agent-ready :D
Article link: economictimes.indiatimes.com…
If there’s a single word in your AI-generated deck that you can’t explain, don’t share it with anyone.
100M+ people in India already use ChatGPT regularly. That’s a massive new discovery and growth channel for Indian businesses.
Excited to deepen our partnership with @OpenAI and bring ChatGPT Ads to India.
The bigger opportunity for me is making this a growth channel that even the smallest business can scale on.
With @Razorpay Engage, AI can do the heavy lifting - launching campaigns, managing them, optimizing performance and scaling what works.
A new growth channel, run by AI and accessible to every business.
People are using ChatGPT extensively to figure out what to buy. Now, the next step in that journey is here.
People don’t just search anymore, they ask questions and do their research. There is nuance and far more depth in the LLM queries versus simple google searches. They ask follow-up questions, compare options and explain what matters to them. This makes for a great place for the right brands to show up at the proverbial right time, which is when consumers are deciding.
So today, we are announcing something I'm quite excited about. @Razorpay Engage is partnering with @OpenAI to bring ChatGPT Ads to Indian businesses.
We're starting with some early partners, @TanishqJewelry, @titanwatches, @ShaadiDotCom, @Palmonas, @AxisMaxLifeIns and more, who are already testing what it means to be discovered inside a conversation instead of a search result.
But every new channel adds complexity for businesses that are already stretched thin. Engage will handle the hard parts. Getting catalogues ready, structuring product feeds, running campaigns, and measuring what actually works, backed by an AI agent and a real team behind it.
Our job has always been to absorb that complexity so founders can try a new way to grow without having to figure it all out themselves. The work with OpenAI already helps founders add Razorpay payments to apps they build with Codex. With Engage, we’re extending that partnership to help businesses find their next customer.
AI led discovery is new territory for everyone, brands and platforms included. We are shaping this frontier actively and help thousands of businsses explore this new channel for their brand.
razorpay.com/engage/chatgpt-…
Really enjoyed this conversation with @aakrit.
We went deep on what it actually takes to become an AI-first company - models, agents, products, culture, and everything in between.
A lot of what we’ve learnt building with AI at Razorpay. And a lot we’re still figuring out.
A few weeks ago at the Global Fintech Fest, I sat down with @harshilmathur, Co-Founder & CEO of Razorpay, to talk about something I’ve been particularly curious about: what does it actually mean to turn a fintech giant into an AI company?
@Razorpay has quietly been doing some fascinating work. Most recently, they announced Vulcan, a foundation model for payments, built from scratch in India to improve fraud detection. And Harshil himself has been one of the more AI-pilled CEOs I’ve come across in the country, hands-on building agents and studying the craft.
We went deep into how they’re building with AI, what it changes across payments, and how he thinks about the technology from first principles.
One of the highest-quality technical conversations I’ve had with a CEO in India.🎙️
@ActivateSignal
Apple Pay is here in India. And it’s already live on Razorpay. 🍎🇮🇳
We’ve always pushed to bring the newest, most seamless ways to pay to India as early as possible.
Starting today, @Apple Pay is live with thousands of businesses on @Razorpay including JioMart, Netmeds, Sephora, Milkbasket and Reliance Digital.
For most businesses using Razorpay Checkout, Apple Pay will show up automatically - with zero changes needed from their side.
No card numbers. No CVV. No OTP. Just authenticate and pay.
More details and early access:
razorpay.com/payments/apple-…
Harshil Mathur retweeted
Yipppieeeee now I can unhook recurring subs of the company from personal cards 🙏
In almost a decade of running a business, only @Razorpay has shown the spirit to enable small businesses to stand on their feet.
Their earlier SBM credit card was such a beautiful product. Get emotional thinking how easy it was.
I have also used their cash advance product (I don't think it exists anymore) a couple of times, seamless quick way to access cash in a tight spot.
Now they are still enabling credit cards for businessess. Was a more intensive process.
Much credit to Vivek & team at Razorpay for patiently hand holding a very irritated me through this lengthy process. I want Razorpay to succeed in enabling small businesses to access credit - hence I persevered. Every success builds a milestone in making processes easier for others. 🙏
@harshilmathur
Building your own foundation model is hard. It shouldn’t require everyone to start from zero.
We’re sharing the technical details behind Vulcan - what we built, how we trained it, and what we learnt along the way.
Hopefully the next team can start a few steps ahead.
A few weeks ago, we launched our foundation model, Vulcan. We’re grateful for the opportunity to present it to our honourable Prime Minister Shri Narendra Modi at GFF ’26.
With Vulcan, we’ve seen an 8-10% improvement in success rate, 5x more disputed transactions & 8x more international card fraud transactions that have been identified while also offering millions of hyper-personalised checkout experiences.
We’re now sharing how we built Vulcan and how it helps us work faster and smarter across our network. The team deserves all the credit for bringing it to life.
@Razorpay
Link: razorpay.com/blog/vulcan-how…
Been using Jev by @typesafeai and it feels like a glimpse of a different AI future.
High-throughput, low-latency intelligence means AI can run inside every loop, reacting in real time.
This is a bigger deal than a better model. It’s a different kind of intelligence.
And once you see it, the idea feels almost obvious in hindsight!
Harshil Mathur retweeted
The introduction of MDR on large-value UPI transactions (i.e., above ₹2,000) is an important step towards strengthening the long-term sustainability of India’s digital payments ecosystem. It will help UPI in continuing to scale, innovate and serve consumers and businesses across the country.
A fair and appropriate distribution of MDR across ecosystem participants will support continued investment in technology, infrastructure and acceptance networks. This, in turn, can enable wider UPI acceptance, deepen the customer base and support sustained growth in transaction volumes.
Importantly, all UPI transactions - P2P and P2M - shall remain free for users. P2M UPI transactions below ₹2,000 will continue to remain free for merchants, while MDR may be levied on merchants for P2M .
RBI remains committed to ensuring that UPI continues to be safe, seamless, affordable, and accessible, while supporting the long-term sustainability and growth of India’s world-class digital payments ecosystem.