Fintech giant Revolut has launched Revolut Research, a dedicated AI research division aimed at building proprietary models rather than relying on third-party solutions. The unit will work alongside academic and technology partners to pioneer new financial services machine learning technology. At its core is PRAGMA, a proprietary foundation model developed with Nvidia designed to understand complex financial behaviour and power fraud detection, credit risk assessment, platform operations, and personalised product recommendations.

Revolut reports early deployments delivered 2.3x better credit risk accuracy, 65% more fraud cases detected, and 41% more relevant product recommendations. Pavel Nesterov, head of AI at Revolut, stated the initiative institutionalises a "build, don't bolt on" philosophy, leveraging data from 80 million customers to create a self-reinforcing loop where growing datasets make models smarter. The unit plans to publish scientific findings, open-source technical frameworks, and engage with the scientific community through conferences and meetups at Revolut's offices.

This move intensifies the AI arms race among Europe's neobanks, following Starling's spending intelligence tool last June, Bunq's upgraded AI assistant in December, and Starling's "agentic" AI assistant in March. Revolut launched its own assistant in April, signalling a broader shift toward native AI infrastructure in digital banking.