Revolut has established an internal, AI-focused research division called Revolut Research. The new unit is tasked with integrating machine learning technologies into the company’s core systems, with particular emphasis on fraud prevention, risk analysis and personalized financial services.
Technical foundation and strategic partner
Revolut Research builds on a foundation model developed in partnership with NVIDIA, known as PRAGMA. The company says it will embed its own AI solutions directly into central systems rather than relying on off‑the‑shelf software. PRAGMA is designed specifically to analyze financial behavior patterns and to support real‑time risk scoring, platform operations and product recommendation engines.
Test results
Anton Repuskó, head of Revolut Research, stated that tests run on historical data already show measurable performance improvements over traditional systems. Reported results include:
- Credit risk analysis: a 2.3× increase in accuracy for identifying default risk.
- Fraud prevention: 65% more frauds were filtered out, with alert accuracy improving by 17%.
- Product recommendations: suggestions for retail and business customers became 41% more relevant.
Applications and future plans
The integrated technology also supports features such as AIR, an assistant embedded in the Revolut app that handles complex financial tasks and is currently available only in the United Kingdom. Revolut Research plans to regularly publish open‑source frameworks and scientific findings going forward.
Why it matters
The move highlights a broader shift in fintech toward building proprietary AI capabilities and integrating them tightly into core banking systems to improve effectiveness and accuracy. The early results Revolut cites suggest meaningful gains in fraud and risk management that could influence product development and customer servicing.
Tags
fintech, artificial intelligence, machine learning, Revolut, research & development, fraud, database, NVIDIA, risk management, credit risk
