The Royal Bank of Canada (RBC) and its AI research arm Borealis AI have teamed with Red Hat and Nvidia to develop an AI computing platform with the goal of improving the client banking experience. The partners are leveraging Red Hat OpenShift and Nvidia’s DGX AI computing systems to power a new private cloud system that they say can perform thousands of simulations and analyze millions of data points in a fraction of the time required previously, allowing “transformative intelligent applications to be brought to market faster.”
RBC will use the in-house, self-service infrastructure to build, deploy and maintain next-generation AI-powered banking applications. In referencing the new DGX hardware, RBC touted the benefits of having a GPU accelerated and containerized platform, built on Red Hat’s OpenShift enterprise-ready Kubernetes distribution, but did not disclose any further technical details.
Nvidia sells DGX platforms based on both its A100 Ampere GPU architecture as well as the previous-generation Volta V100 parts. These can be networked into DGX Superpods to create very large clusters, such as Selene, Nvidia’s in-house supercomputer that delivers 27.5 double-precision Linpack petaflops and more than 1 exaflops of “AI performance.”
A collaborative announcement notes that RBC’s new platform “has already improved trading execution and insights, helped reduce client calls and has resulted in faster delivery of new applications for RBC clients, and has the potential to benefit the AI industry in Canada, beyond RBC and financial services.”
The project underscores the growing importance of HPC and AI for the banking industry.
“Modern AI cannot exist without access to high performance computing. This collaboration means that we can conduct research at scale, and deploy machine learning applications in production with improved efficiency and speed to market,” observes Foteini Agrafioti, chief science officer, RBC and head of Borealis AI.
Canada’s largest bank, RBC founded the Borealis AI research institute in 2016 to focus on advanced AI opportunities. Earlier this year, the bank filed patents for several artificial-intelligence inventions, including one that uses a recurrent neural network to predict when clients will make major purchases, according to reporting in the Financial Post.