Razorpay has launched a transformer-based AI foundation model Vulcan built to expedite digital payments. The IPO-bound startup claims that the model can make every digital payment more reliable and safer by scoring routes in real-time and flagging fraud that becomes visible across multiple merchants after it is in production.

Built with NVIDIA and AWS technology, Vulcan combines Razorpay’s payments data with NVIDIA’s accelerated computing and AWS’ cloud infrastructure. 

Vulcan is a proprietary, ground-up model, with both its architecture and training data owned by Razorpay. The startup frames it not as an LLM, which understands text, but as a system that learns the movement of the money. 

The model learns from the entire payments ecosystem’s data points and keeps improving with every transaction it processes, instead of solving one narrow problem at a time. 

The startup claims that Vulcan is trained on approximately 3 Tn data points across 4 Bn payments and learns from roughly 3,000 signals per transaction. With this, it understands the complex movement of the financial journey at scale. 

Razorpay reports early results from deploying Vulcan’s components on live transactions. According to the company, the model has delivered:

“Every payment teaches the system something that makes the next payment better. That’s what makes this feel less like a product launch, and more like the starting point for how payments in India keep getting better on their own, for years to come,” CEO Harshil Mathur said.

While Razorpay claims to be the first Indian player to introduce this offering, global incumbent Stripe also has previously launched a payments foundation model to analyse payments patterns, trained on tens of billions of transactions, to learn and boost performance and help its customers grow revenue.

Source link