Walmart Inc.’s marketing strategy has focused on affordable prices and seamless customer service, and that now extends to its machine learning platform, Element.

The centralized, cloud-native service allows Walmart to harness the power of large language models for all of its online and physical commerce, according to Anil Madan (pictured), senior vice president of cloud and data platforms at Walmart Global Tech.

“We manage from a data life cycle, thousands and thousands of data sources. The data lifecycle is data moving in more in a raw format, coming into a central lake. It goes through certain transformations and quality to create a data catalog. That data catalog helps very much power every aspect of our analytics and machine learning,” he said, emphasizing how Element has changed the game for Walmart’s cloud operations. “The key elements of our enterprise data lake is just how we build this in a centralized form by still respecting the foundational elements of security, data sovereignty.”

Madan spoke with theCUBE Research’s Paul Gillin at the Supercloud 7: Get Ready for the Next Data Platform event, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed how Element powers Walmart’s enterprise and its best-of-breed strategy.

The triple model underlying Element

Walmart uses a hybrid cloud environment, with Element powering analytical processing, machine learning and data management across private and public clouds to create what Madan describes as a seamless, “omnichannel” experience.

“What we have is the hybrid multicloud strategy, which basically helps us seamlessly integrate and run cloud-agnostic workloads, both application workloads as well as ML workloads,” he said. “When we talk about the triplet [model], we have two public cloud providers and a private cloud coming in a symmetrical fashion so that we can basically run hybrid multicloud workloads at scale.”

Element’s “triplet model” is an umbrella for Walmart’s open-source OneOps Cloud Management Platform, its Walmart Cloud Native Platform, which means the company can move applications between private and public clouds and a data abstraction layer that allows engineers to move data. These cloud providers are then replicated in three regions: West, Central and East.

What [Element’s triplet model] provides basically is our on-demand infrastructure where we can take different compute types, whether CPU, GPU, TPUs, and run them portably across these different cloud providers. We also have a very, very matured MLOps deployment framework, which basically helps us deploy these workloads in minutes versus days,” Madan said. “Those things in conjunction now are helping power all our generative AI workloads as well, because now we can interoperate different kinds of LLMs … to run them seamlessly across this triplet architecture.”

The abstraction layers deployed across the triplet architecture also let engineers more easily train and scale artificial intelligence models. As a result, Walmart’s AI development is moving quickly without the company having to lock in to any one vendor.

“You have the cloud management abstraction layer with OneOps, you have the workload management layer with our WCNP, and then you have the data management layer in how we provide SDKs to our application developers,” Madan said. “A combination of these abstraction layers basically help us then provide those vendor-agnostic, best-of-breed technologies, which can basically help them swap with speed if we need to while still giv[ing] them the ability to deploy these at scale fairly quickly and rapidly in this hybrid multicloud.” 

Tailoring platform to ethos

Walmart has avoided committing to a single vendor in its development of Element, with the goal of maintaining its “Every Day Low Cost” business model. Open-source software also helps the company save money on its operations, and for its customers.

“The payoffs [on Element] are across all aspects. And the developer productivity is just a huge, huge payoff,” Madan said. “The biggest benefit to these application engineers and data scientists is they can rapidly train their workloads at scale. They’re not going and sourcing data to create all of that because now all of that has been solved for them in a common enterprise data lake. They just go, they pick their algorithm or what they want and they basically go and they’re operational.”

The rapid AI development Element has fostered does not come at the cost of security, according to Madan. Walmart has implemented its own governance layer to detect AI hallucinations and mitigate security risks in an effort to maintain its status as a trusted retailer.

“Everything at Walmart is anchored around our mission and purpose … we are a people-led, tech-powered, omnichannel retailer, and our whole purpose and goal in life is to save money so that we can help our customers live better,” Madan said. “Security is a critical underpinning to everything we do, so we make sure it’s how data gets used, consumed, who has authorization, that becomes a critical piece into our guiding principle.”

Stay tuned for the complete video interview, part of SiliconANGLE’s and theCUBE Research’s coverage of the Supercloud 7: Get Ready for the Next Data Platform event.

Photo: SiliconANGLE