Data center infrastructure supplier Nutanix Inc. today is jumping into the generative pre trained transformer fray with GPT-in-a-Box, which it describes as a toolkit for companies that are looking to jumpstart artificial intelligence and machine learning projects without risking their data or budget.

Nutanix described the new offering as a full-stack, software-defined artificial intelligence-ready platform that includes all the elements needed to build AI-ready infrastructure. It includes services to help organizations size and configure hardware and software infrastructure to support a curated set of large language models using popular open-source AI and machine learning operations frameworks on the Nutanix Cloud Platform. Customers can procure AI-ready infrastructure to fine-tune and run generative GPT models, including LLMs.

“AI applications are new and introduce performance considerations that are different from traditional applications,” said Lee Caswell, senior vice president of product and solutions marketing at Nutanix in emailed comments. “Hardware to develop large language models, for example, can be quite different from hardware best suited to run AI inferencing. Similarly, different AI software workflows can require different performance needs between, say, a natural language processing and a video analytics app.”

The company said the offering is aimed at companies that want to begin working with AI but are concerned about intellectual property, compliance and privacy issues. They also struggle with how to best support machine learning administrators and data scientists and the risk that costs could rise out of control.

Security concerns

McKinsey & Co. reported that cybersecurity, regulatory compliance, privacy and model explainability are the top risks that hold organizations back from investing in AI and that there has been little perceived mitigation of any of those risks since 2019. “There is also a consideration around large upfront AI investment costs which can stall AI and ML initiatives,” Caswell said. “We are offering a service to help customers optimize their infrastructure with an eye to preserving data privacy, governance, and security.”

Nutanix GPT-in-a-Box is described as a ready-to-use customer-controlled AI infrastructure package for edge or data center deployment that includes a full set of security and data protection offerings.

It includes Nutanix’s cloud infrastructure, file and object storage, the Nutanix AHV hypervisor and Kubernetes platform with Nvidia Corp.’s graphics processing unit acceleration, services to help users size their cluster and deploy open-source deep learning and MLOps frameworks, an inference server and a curated set of large language models, including Llama2, Falcon LLM and MosaicML.

The company said data scientists can immediately use the supplied models with their choice of applications as well as with an enhanced terminal or command language interface. The platform can also be leveraged to run other GPT models as well as to fine-tune models leveraging internal data hosted on the included file and object services.

Pricing wasn’t specified.

Photo: Nutanix