Revolut has launched this Tuesday a new division focused on research, integrated into its Artificial Intelligence (AI) Department and named Revolut Research, with the aim of leading the development of next-generation machine learning architectures applied to the financial sector on a global scale.
According to the firm, this unit will work alongside leading academic and technological institutions and "lays the unified foundations for all exclusive AI initiatives and solutions of Revolut." "If we want to lead the future of smart banking, we cannot rely on third-party solutions," emphasized Revolut's AI head, Pavel Nesterov.
For the entity, Revolut Research is "the engine that drives Pragma," the proprietary foundational model developed alongside Nvidia. "This model has been specifically designed to decipher complex financial behaviors and optimize real-time risk assessment, platform operations, and personalized product recommendations," indicated the neobank.
With this move, Revolut claims it is rethinking its entire AI infrastructure on a single foundation, "equipped with a common layer of behavioral intelligence."
Regarding the performance of Pragma, the company has revealed that this technology multiplies by 2.3 the accuracy in detecting the risk of default, which translates into a more solid assessment of users' solvency and more tailored credit granting decisions.
In the area of combating fraud, Pragma has allowed for the identification of 65% more fraudulent operations, with a 17% increase in the accuracy of alerts. As for hyper-personalization, Revolut claims that product suggestions are now 41% more relevant to the needs and preferences of its customers, both individuals and businesses.
On the construction of Pragma, the company highlights as a key element its global database, which allows it to process billions of cross-border transactions from over 80 million customers and examine different patterns of financial behavior in real time.
This information feeds the models of Revolut Research, generating a "constant improvement loop." "The more information is processed, the more accurate the models become in stopping fraud, measuring risks, and anticipating user needs, thus consolidating an exclusive analytical advantage," the entity has emphasized.
"With Pragma, we are developing a unique and unified foundational model, capable of understanding the true nuances of financial behavior in real time. Technology is in the DNA of Revolut, and thanks to collaboration with global technology leaders, this division is designing exclusive capabilities that differentiate us from traditional banking," concluded the head of Revolut Research, Anton Repushko.
In the coming months, Revolut expects the Research unit to continue expanding its scientific activity through the periodic publication of results and the launch of new open-source technical frameworks.
Likewise, the company will deepen its collaboration with the international academic community through its presence in specialized forums such as Nvidia's GTC in Berlin, scheduled for October, and the ICAIF, which will take place in November. At the same time, Revolut Research will organize quarterly meetings in its offices to promote knowledge exchange and cooperation with researchers from different countries.