Revolut Launches Dedicated AI Research Division to Build Native Foundation Model for Banking

2 Sources

Share

Revolut has established Revolut Research, a dedicated AI research division within its AI department, to develop Pragma—a foundation model trained on data from 80 million customers across 40 markets. Built in partnership with Nvidia, Pragma has demonstrated a 2.3x uplift in identifying credit default risk and 65% improvement in fraud detection compared to legacy systems.

News article

Revolut Research Establishes Dedicated AI Division for Native Foundation Model Development

Revolut has launched Revolut Research, a specialized division within its broader AI department focused on advancing machine learning architecture across its fintech platform

1

. The neobank's strategic move signals a decisive shift toward building native AI capabilities rather than relying on third-party solutions, as the company aims to reshape intelligent banking through proprietary technology.

The Research unit serves as the AI engine driving Pragma, Revolut's foundation model developed in partnership with Nvidia

2

. Unlike traditional approaches that treat each transaction as an isolated task, Pragma unifies financial behaviors into one connected system, processing data from more than 80 million consumers across 40 markets to build patterns of financial behaviors in real time

1

.

Pragma Foundation Model Delivers Measurable Performance Gains in Risk and Fraud Detection

Early deployments of Pragma on historical data have revealed substantial performance improvements over legacy systems. The foundation model achieved a 2.3x uplift in identifying credit default risk, caught 65% more fraud cases, and delivered 41% more relevant product recommendations across retail and business accounts

2

. These metrics demonstrate how customer transaction data, when processed through sophisticated machine learning architecture, can significantly enhance real-time risk assessment capabilities.

Pavel Nesterov, head of AI at Revolut, emphasized the strategic importance of this approach: "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"

1

. By training native foundation models on global operational data, Revolut aims to give engineering teams an unprecedented engine to deploy smarter features faster and eliminate systemic friction

2

.

Building Fintech Architecture Through Collaboration and Open-Source Initiatives

Revolut Research plans to publish findings from its scientific research and open-source technical frameworks, integrating its team into the wider scientific community through conferences and quarterly meet-ups

1

. This commitment to transparency positions Revolut to refine the intelligence layer underpinning Pragma while contributing to broader advancements in AI research

2

.

Anton Repushko stated that Revolut Research was established to "responsibly build financial intelligence at its deepest layer, rather than patching together narrow, specialised models"

1

. The division's work with Nvidia focuses on developing a single, unified foundation model capable of understanding the true nuance of financial behaviors in real time, setting Revolut apart from traditional banks.

What This Means for Fintech and AI-Driven Banking

As Revolut processes billions of cross-border transactions from its 80 million customers, the dataset feeding into Revolut Research's models grows continuously. This creates a compounding advantage: as the dataset evolves, the models become exponentially smarter at fraud detection, evaluating risk, and predicting user needs

2

. The platform's ability to detect fraud while enabling various customer services through one unified system represents a fundamental shift in how fintech architecture can be designed.

Watch for how Revolut's open-source contributions influence the broader AI department strategies across financial services, and whether the performance gains demonstrated by Pragma push other neobanks to invest in proprietary foundation models rather than off-the-shelf solutions. The 2.3x improvement in credit default risk identification alone could reshape lending practices, while the 65% boost in fraud cases caught may set new industry benchmarks for real-time risk assessment.

Today's Top Stories

© 2026 TheOutpost.AI All rights reserved