Nvidia AI Enterprise 3.0 adds new application workflows, partners with Deutsche Bank

Modern artificial intelligence (AI) workloads need both hardware and software for enterprises to recognize the full benefits.

Today, Nvidia is pushing forward on the software front, announcing a new partnership with global financial services firm Deutsche Bank to help enable more advanced AI capabilities across multiple use cases at the bank. Nvidia is also releasing its Nvidia AI Enterprise 3.0 today, which brings new software capabilities — including application workflows — to help organizations like Deutsche Bank more effectively build and deploy AI-driven applications.

Nvidia AI Enterprise debuted in 2021 and has been iterated on with a steady release cadence ever since. In July of this year, the 2.1 release came out, with a focus on updating open-source tools for machine learning (ML), including PyTorch. A core feature of Nvidia AI Enterprise is that it can run in an enterprise’s own data center, on the cloud, or in a hybrid approach across both types of deployments.

“Nvidia’s GPU hardware really, in some ways, opened up the field of AI, because the AI algorithms are able to run so much faster on GPUs,” Manuvir Das, VP of enterprise computing at Nvidia, said during a press briefing. “What we quickly learned at Nvidia is that in order to benefit from accelerated computing, the workloads, use cases, tools and frameworks have to be adapted and retargeted to GPUs.”

There is a lot of AI software that enterprises might need

Das explained that Nvidia Enterprise is a large body of AI software.

That software includes the core ML libraries with TensorFlow and PyTorch. It also includes the Rapids data processing library for Python developers. The Triton library for AI inference is another core element of Nvidia Enterprise. 

The Riva speech AI as well as Morpheus security frameworks are also part of Nvidia Enterprise AI.

“AI is complicated because you have lots of these open-source pieces that come into play that one has to sort of stitch together,” Das said. “Nvidia Enterprise really provides enterprise customers a singular platform where they know that all these pieces are integrated together, tested together, and all of it’s available for enterprises to use.”

Application workflows land in Nvidia Enterprise 3.0

The big update in Nvidia Enterprise AI 3.0 is not in any one framework or tool, but rather in a new set of capabilities known as application workflows.

“Workflows are packages of software and models for use cases that are much closer to the problems that the customer is working to solve,” Das said.

One such example is digital fingerprinting for cybersecurity use cases, where AI is used to build a model of the behavior of every employee in your company. The model creates a digital fingerprint of the user, making it possible to identify when there is any deviation in behavior, which could potentially be an indicator of an impersonation attack.

Das explained that the workflows are a combination of software frameworks and pretrained models. The workflows can also be deployed using Kubernetes Helm Charts, which provide a mechanism for organizations to deploy applications into cloud-native container environments.

The pretrained models are also available in an unencrypted format, which Das said makes it possible for organizations to analyze on their own. Allowing organizations access to unencrypted models also helps to support explainable AI efforts, such that organizations will be able to understand how the models work.

Ist ein deutsches AI? Deutsche Bank embraces Nvidia Enterprise AI

Among the organizations that are set to benefit from Nvidia Enterprise AI 3.0 is Deutsche Bank, which announced today that it is partnering with Nvidia to help advance the bank’s AI effort.

“This is a commitment on our side to actually make AI an integral part of the way we function across the bank,” Gil Perez, chief innovation officer and global head of cloud and innovation network at Deutsche Bank, said during a press briefing.

Perez said that, to date, Deutsche Bank has been using AI in a lot of what he referred to as traditional use cases. Those use cases include matching names and payment categorization. Now, together with Nvidia, he said that Deutsche Bank is looking to do more real-time enablement, using AI to help improve business outcomes.

“The key difference between where we are today and the future, is that today we are doing things in a batch way, and tomorrow it’s really seamless and real time,” Perez said.