AI Lab Could Make WWT ‘Central Hub’ Of Customer Innovation

POCs, education workshops and on-demand capabilities from the lab are expected by the end of the first quarter.

World Wide Technology plans to have about 12 artificial intelligence lab environments built out by the end of 2024’s first quarter as part of an AI “proving ground” for validating software, use cases and components of an AI product or service.

That’s according to Matt Halcomb, principal architect at St. Louis-based WWT – No. 9 on CRN’s 2023 Solution Provider 500 – who gave more details on WWT’s new project to CRN in an interview.

He envisions WWT as the center hub of a multi-vendor wheel bringing technologies together for AI projects.

“How do we help a customer who has to move from having an executive team throwing ideas around a board table into leveraging data to get to data-driven outcomes and business decisions that they’re doing?” he said. “Generative AI is a great component of that.”

[RELATED: WWT Reveals $500M Spend Planned For Helping Customers Adopt AI]

WWT AI Lab

Proofs of concept (POCs), education workshops and on-demand capabilities from the lab – part of WWT’s investment of $500 million over three years in AI – are expected by the end of the first quarter, he said. More on-demand capabilities will come in the second quarter, allowing customers to schedule a lab environment for a week or two.

WWT’s website will allow some self-service for customers, but data scientists and engineers will be available to help customers with projects and education, Halcomb said.

WWT is still working out the pricing for the lab environments. But some on-demand components will be free depending “on the complexity of the ask.”

About a dozen customers already have access to the handful of environments built out. One use case, so far, is POCs for purchased hardware that hasn’t arrived yet, experimenting with integration options, he said.

“A lot of organizations do really good at building infrastructure, but they don’t do such a good job of enabling the end user to leverage the infrastructure,” he said. “And that’s the key component to that return on investment.”

The solution provider is already about finished with a handful of environments and it has customers already trying them out.

WWT has five data centers and its Advanced Technology Center (ATC) to leverage for the project. Eventually WWT will integrate its ATC and infrastructure into a cloud environment and it will create a cloud-adjacent edge location, Halcomb said.

The lab will also help customers understand what technologies they really need to adopt. Sometimes customers won’t need cloud – such as using cloud-adjacent and edge environments to enable computer vision.

WWT will also evaluate new software and hardware vendors to ensure they can deliver on what they’ve advertised to a customer, he said.

“We’re more than willing to take a look at a number of different technologies and solutions and help a customer make intelligent decisions for themselves – quit guessing at things,” he said. “The technology we’re looking at and the technology they’re looking to implement, they want to do it in a hurry, but we don’t want them to be in a hurry to do the wrong thing because it has a huge impact on their budget. It has a huge impact on their carbon footprint.”

Mass adoption of cloud and then realizing how expensive cloud can be also plays into the lab’s usefulness, Halcomb said. Sometimes customers want to buy more servers and workstations than they need for a project.

“Let’s take a look at what your needs are today,” he said. “But let’s make sure that we architect your environment in a scalable fashion so that as your consumption goes up, you can then begin consuming those components in a business-intelligent way.”

WWT will leverage partnerships with a variety of software and hardware vendors so that customers can use preferred partners and experiment. WWT will have reference architectures from Hewlett Packard Enterprise (HPE), Dell, Nvidia and other vendors.

The solution provider can spin up a node for customers with a specific graphics processing unit (GPU), allow the customer to validate a vendor, and then pull that away and move to the next one, he said.

WWT is almost done with an automated system for pre-built ISOs, allowing end users to select the vendor processor, server, a pre-trained large language model (LLM) and other factors, Halcomb said. Customers can compare different parameter models on different hardware.

Halcomb is also working on a training lab with some Nvidia DGX systems that will allow users with role-based access to train a model and bring it into a Nvidia Omniverse environment.

Customers will have “the ability to mix and match components at different layers of the stack that’s important to them,” Halcomb said.

WWT will have the ability to run parallel environments for customers who run into brick walls and want to pivot fast.

To WWT’s vendor partners, Halcomb asks that they stay engaged and close to make the lab environments the best they can be.

“Let’s not be afraid to align with multiple partners because it is a multi-stack environment, and I have yet to see any one organization that truly can take on every single workload,” he said.