Modulr to create AI models purpose-built for payments

London, UK | 14 October 2026: Modulr, the payments automation platform built to scale, today announced it is developing purpose-built AI open models powered by NVIDIA for scaling payment operations. The models run inside Modulr’s growing network of AI agents that already operates across its platform and are part of a wider programme to bring agentic intelligence to payments.

Payments provide a challenging environment for AI innovation due to real-time, high volume money movement that is regulated, with no room for error. For platforms like Modulr’s, which connects directly into UK and European payments rails alongside card networks and cross-border payments, the infrastructure must be resilient, highly available, accurate, and scalable. 

To address this, Modulr’s is building a family of specialised models, each tuned to a distinct part of payment operations – from the movement and reconciliation of funds to risk, resilience, and the processes behind every transaction – keeping a person in review before any action is taken. The models are drawn from NVIDIA’s Nemotron-3-Ultra-550B open model, running on NVIDIA accelerated computing, and fine-tuned on Modulr’s own standards and payment expertise. 

The models are self-hosted and owned by Modulr, giving the transparency, control, and data residency needed to run safely at scale. The first two models in this family, Scout and Soldier, are focused on engineering and security and trained on Modulr's own standards and payment expertise, adding an intelligent layer over its money movement software. Initial signals show that the open models achieve the same level of performance as frontier models with an accuracy of over 90% while running on open, efficient infrastructure, at a fraction of the cost. 

For Modulr’s customers, it also means a payments automation platform that runs with greater accuracy and resilience as it grows, with plans to extend capabilities to customers directly in future.

Vivek Awasthi, Chief Technology Officer, Modulr, said: “Payments are complex by nature, and need purpose-built AI solutions that can support layered infrastructure and scale safely. We are deploying specialised models that understand payments the way our own experts do, with a person accountable for every action. Working with NVIDIA and its Nemotron models lets us run open models we control ourselves that are efficient, data-driven, and transparent.”

Georgios Kolovos, Payments & FinTech Leader EMEA, NVIDIA, added: “Complex sectors like payments need agentic software they can run efficiently and govern with confidence. With NVIDIA open models and accelerated computing, Modulr is building payments intelligence that is production-ready from day one.”