Revolut has established Revolut Research, a dedicated division within its wider AI Department, created to pioneer next-generation machine learning architecture across global financial services. The unit will build alongside leading academic and technology institutions, and provides a unified foundation for the company’s proprietary AI deployments and machine learning initiatives.
The announcement, made on 25 August 2026, lands as the wider banking industry runs into a technological ceiling. Legacy institutions have spent years retrofitting decades-old infrastructure with off-the-shelf software wrappers, an approach that creates soaring operational drag for the organisations themselves and leaves customers with clunky applications and interfaces, and basic search-bar chatbots.
Revolut, which serves more than 80 million retail customers, has taken the opposite route. Intelligence developed in-house is embedded directly into the core financial engine, turning local market data into a compounding global advantage. The new division formalises that method and gives it a permanent home.
Building in-house
Revolut Research operates as the institutional engine behind PRAGMA, the company’s proprietary foundation model, developed in collaboration with NVIDIA. The model has been engineered specifically to decode complex financial behaviours and to power real-time risk assessment, platform operations and tailored product recommendations. Its arrival marks a rebuild of the entire AI stack around a single foundation with one shared behavioural intelligence layer.
That structural choice carries commercial weight as well as technical merit. Narrow, specialised models each demand their own data pipelines, monitoring and maintenance, and the operational load grows with every addition to the estate. A unified foundation allows engineering teams to draw on the same behavioural understanding for fraud, credit and personalisation alike, which shortens the distance between an idea and a live product feature.
“To lead the future of intelligent banking, you cannot rely on third-party blueprints. We have launched Revolut Research to institutionalise our ‘build, don’t bolt on’ philosophy. By training native foundation models on our global operational data, we are giving our engineering teams an unprecedented engine to deploy smarter features faster, eliminate systemic friction, and give our customers a safer, radically better financial experience.”
Pavel Nesterov Head of AI Revolut
Early results from PRAGMA
Anton Repushko, Head of Revolut Research, has already presented key findings at multiple high-profile international conferences this year, including ICML in Seoul. Early deployments of PRAGMA on historical data have indicated major performance gains over legacy baselines, and the numbers speak to the operational value of a shared intelligence layer.
Credit risk assessment has proved 2.3x more accurate at identifying default risk, which widens the pool of customers who can be safely served while reducing exposure on the accounts that carry genuine danger. Fraud prevention has recorded 65% more cases caught alongside 17% greater precision in alerts, a combination that matters because higher detection usually arrives with more false positives and more frustrated customers. Personalisation has improved by 41% in the relevance of product recommendations across retail and business accounts, turning the app from a place where products are listed into one where the right product surfaces at the right moment.
The value of 40 markets
The advantage rests on an unmatched global dataset. Serving 80 million customers across more than 40 markets, the company processes billions of cross-border transactions and diverse financial behaviours in real time, feeding a volume and variety of signal that few institutions can assemble.
This data flows into Revolut Research’s models and creates a self-reinforcing loop, establishing what the company describes as a new language of financial behaviours. As the dataset grows, the models become exponentially smarter at detecting fraud, evaluating risk and predicting user needs. The result is a proprietary intelligence advantage that traditionally structured banks find difficult to match, since it cannot be licensed, purchased or replicated without the underlying customer base and the global footprint that generates it.
Security and the everyday experience
Beyond PRAGMA, advanced security models review nearly one billion transactions every month to stop fraudulent activity proactively. Prevention on that scale depends on models that understand normal behaviour well enough to recognise the abnormal, which is precisely the capability a unified foundation is designed to deliver.
Connected engines power front-end capabilities including AIR, the in-app assistant that executes complex financial tasks in a single step and is currently available in the UK only. The same intelligence that scores risk in the background is therefore doing visible work for the customer, which shows how research investment translates into features people use every day.
Publishing the work
Revolut Research plans to regularly publish its scientific findings and open-source technical frameworks, a policy that positions the division as a contributor to the field as well as a beneficiary of it. The team will engage with the broader scientific community through international conferences including NVIDIA GTC Berlin in October and ICAIF in November, and will host quarterly meet-ups for the scientific community at the company’s offices.
Open publication and regular community engagement serve a practical purpose. Financial machine learning talent gravitates towards organisations where work can be shared and tested in public, and a division that publishes builds a recruitment advantage alongside its technical one.
“Revolut Research has been established to responsibly build financial intelligence at its deepest layer, rather than patching together narrow, specialised models. In PRAGMA, we are developing a single, unified foundation model capable of understanding the true nuance of financial behaviour in real time. Technology is in Revolut’s DNA, and by collaborating with global tech leaders, this division is engineering proprietary capabilities that set us apart from traditional banks.”
Anton Repushko Head of Revolut Research Revolut