Razorpay Launches AI Foundation Model for Payments

Indian fintech company Razorpay has launched Vulcan, an AI foundation model aimed at improving payment processing. The company describes Vulcan as India's first transformer-based AI foundation model for the payments industry.

The model was developed in partnership with Nvidia and Amazon Web Services (AWS). It is designed to optimize payment routing, detect fraud and risks, and enhance customer experiences through personalization. Several customers, including Blinkit, Bachatt, and redBus, have reportedly begun using some of Vulcan's capabilities in live transactions.

Vulcan has been trained on an extensive dataset, reportedly comprising 4 billion payments and 3 trillion data points. The model analyzes approximately 3,000 signals per transaction, drawing on data across merchants, payment instruments, issuers, and gateways. Razorpay claims the model has achieved an 8-10% improvement in payment success rates and an eight-fold increase in international card fraud detection.

Razorpay is not charging merchants an additional fee for using Vulcan. Founder Harshil Mathur has likened Vulcan to large language models (LLMs), stating that it understands money movement similarly to how LLMs understand language. The company intends to expand Vulcan's capabilities to authentication and lending.