Revolut, the global financial services firm serving more than 80 million retail customers, has announced the launch of Revolut Research, a new division within its broader AI Department focused on advancing machine learning architecture for financial services.
The unit will serve as the institutional engine behind PRAGMA, Revolut’s proprietary foundation model developed in collaboration with NVIDIA. PRAGMA is designed to support real-time risk assessment, platform operations, and personalised product recommendations by analysing complex financial behaviours across Revolut’s customer base.
According to the company, early testing of PRAGMA against historical data has shown notable improvements over legacy systems, including 2.3 times higher accuracy in identifying credit default risk, a 65 percent increase in fraud cases caught with 17 percent greater alert precision, and 41 percent more relevant product recommendations for retail and business customers.
Revolut said the division reflects a broader strategy of building AI capabilities internally rather than relying on third party software. Pavel Nesterov, Head of AI at Revolut, said the company is aiming to “institutionalise” a build first approach by training models on its own operational data.
Anton Repushko, Head of Revolut Research, has presented findings at international conferences including ICML in Seoul. The division plans to publish research and open source technical frameworks going forward, with further appearances expected at NVIDIA GTC Berlin in October and ICAIF in November, alongside quarterly meet ups at Revolut’s offices.
Revolut operates in more than 40 markets, processing billions of transactions used to train its models.
