Razorpay unveils Vulcan, a specialised AI model developed with NVIDIA and AWS, poised to revolutionise digital payments by enhancing security, reliability, and efficiency through innovative real-time payment behaviour analysis.
Razorpay has introduced Vulcan, a transformer-based artificial intelligence model built specifically for digital payments, in a move the fintech company says will make online transactions more reliable, secure and efficient. Developed with NVIDIA and AWS technology, the system is designed to assess payment behaviour in real time and help decide how a transaction should be handled at each stage.
Unlike general-purpose large language models, Vulcan has been trained to interpret payment data rather than text. Razorpay says its architecture and training data are proprietary, and that the model learns from money moving across its payments network. The company says it was trained on nearly 3 trillion data points from more than 4 billion payments and examines about 3,000 signals per transaction.
Those signals are used for tasks including routing, fraud detection, risk assessment and checkout optimisation. Razorpay says early use of parts of Vulcan has lifted payment success rates by about 8% to 10%, while also flagging as much as eight times more international card fraud. It said the system identified five times as many fraudulent or disputed transactions without increasing alerts.
The model was developed after Razorpay studied payment behaviour among about 1.5 million shoppers and more than 51,000 businesses. That work pointed to friction across both large cities and smaller markets, underscoring the challenge of making digital payments more predictable in India’s fragmented ecosystem, where consumers and merchants use a mix of banks, cards, UPI apps and other payment methods.
Razorpay also plans to extend Vulcan into lending and payment authentication, suggesting it could become a broader layer across the company’s financial services stack. The launch reflects a wider shift in India’s fintech sector, where artificial intelligence is being used not just for customer support, but in the core infrastructure that determines whether payments go through, whether fraud is caught and how risk is managed. The company’s existing tools for success-rate analytics, optimiser-driven routing, authentication and anomaly detection point to a longer-running push in that direction.
Disclaimer: This article is intended to inform and educate, not to recommend or endorse any financial product, investment or strategy. Please consider your own financial circumstances and seek professional advice where appropriate before making financial decisions.
