New Delhi, Aug 18: Fintech company Razorpay has launched Vulcan, an artificial intelligence-powered foundation model designed specifically for the digital payments ecosystem, marking a major step in the use of AI to improve payment reliability, fraud detection and the online checkout experience.
Built with technology support from NVIDIA and AWS, Vulcan has been developed specifically to understand payment behaviour and transaction patterns. Razorpay said the model is designed to handle the complexity of India’s rapidly expanding digital payments ecosystem and make real-time decisions during transactions.
The model has been trained on nearly 3 trillion data points from around 4 billion payments and evaluates about 3,000 signals for each transaction. These signals can help determine the most effective payment route, identify potentially fraudulent activity and improve the chances of a transaction being completed successfully.
For customers, one of the most visible benefits could be fewer failed or delayed payments. For merchants, higher payment success rates can directly translate into fewer abandoned purchases and better revenue conversion.
Razorpay said early results from Vulcan have shown an 8-10 per cent improvement in payment success rates. The company has also reported improvements in fraud detection without a corresponding increase in unnecessary alerts.
The model is already being used in live payment environments by businesses including Blinkit, Bachatt and redBus, according to the company.
Vulcan is also being used to make digital checkouts more personalised. Through Razorpay’s Magic Checkout platform, the system can identify a customer’s preferred UPI application and present it during checkout. The company said this has resulted in 40 per cent more shoppers being shown their preferred UPI app, supporting an additional 1-2 lakh purchases each month.
Unlike general-purpose AI models that are primarily trained to understand language and text, Vulcan has been developed around financial transaction data and payment behaviour. This allows it to focus on the specific challenges faced by payment platforms, including routing transactions efficiently, identifying unusual activity and managing payment risks.
Security and data protection are also central to the system. Razorpay said personally identifiable information is removed from transaction data before it is used by the model, while the system operates within the company’s private infrastructure in India.
The launch comes as digital payments continue to become a central part of everyday commerce in India. From small retailers to large online businesses, companies increasingly depend on payment systems that can process transactions quickly while keeping fraud and failures under control.
Even small improvements in payment success rates can have a significant commercial impact when applied across millions of transactions. A more effective payment-routing system can help businesses recover transactions that might otherwise fail, while improved fraud detection can reduce potential financial losses.
Razorpay plans to expand the use of Vulcan beyond payment routing and fraud detection, with potential applications in areas such as authentication, risk management and lending.
The launch reflects a wider shift in India’s fintech sector, where AI is increasingly being deployed behind the scenes to improve the infrastructure that supports digital commerce.
For consumers, the objective is a smoother and more reliable payment experience. For businesses, the technology could mean fewer failed transactions, stronger fraud protection and improved conversion at checkout.
