Explore AI innovations in data centers of the future with Dave Vellante, Co-Founder and Co-CEO of SiliconANGLE Media. He hosts Sravana Karnati, Executive Vice President of Global Technology Platforms at Walmart, on theCUBE's coverage of AI factories: data centers of the future. This insightful discussion explores Walmart’s evolving AI strategies and innovations, providing a unique perspective on how one of the world's largest companies integrates cutting-edge technology into its global operations.
In this episode, Karnati shares their expertise and insights on Walmart’s approach to AI and cloud infrastructure. They explain the triplet model strategy for data centers, emphasizing the importance of a multi-region hybrid cloud environment. The discussion also highlights the Walmart Cloud Native Platform, which enables consistent development and deployment across varied cloud platforms, as well as the transformative impact of generative AI technologies on Walmart's operations.
Key takeaways from this conversation include the development of WIBEY, Walmart's super-agent platform enhancing developer productivity by eightfold, and the integration of agentic and generative AI technologies. Karnati contrasts deterministic microservices with the more flexible context-aware functionality provided by WIBEY. The discussion offers valuable insights into AI’s potential to revolutionize developer workflows and streamline business processes, according to Karnati and theCUBE analysts.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: The AI Factory - Data Center of the Future. If you don’t think you received an email check your
spam folder.
Sign in to AI Factories - Data Centers of the Future.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Register for AI Factories - Data Centers of the Future
Please fill out the information below. You will receive an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for AI Factories - Data Centers of the Future.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: The AI Factory - Data Center of the Future. If you don’t think you received an email check your
spam folder.
Sign in to AI Factories - Data Centers of the Future.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: The AI Factory - Data Center of the Future
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: The AI Factory - Data Center of the Future. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Sravana Karnati, Walmart
Explore AI innovations in data centers of the future with Dave Vellante, Co-Founder and Co-CEO of SiliconANGLE Media. He hosts Sravana Karnati, Executive Vice President of Global Technology Platforms at Walmart, on theCUBE's coverage of AI factories: data centers of the future. This insightful discussion explores Walmart’s evolving AI strategies and innovations, providing a unique perspective on how one of the world's largest companies integrates cutting-edge technology into its global operations.
In this episode, Karnati shares their expertise and insights on Walmart’s approach to AI and cloud infrastructure. They explain the triplet model strategy for data centers, emphasizing the importance of a multi-region hybrid cloud environment. The discussion also highlights the Walmart Cloud Native Platform, which enables consistent development and deployment across varied cloud platforms, as well as the transformative impact of generative AI technologies on Walmart's operations.
Key takeaways from this conversation include the development of WIBEY, Walmart's super-agent platform enhancing developer productivity by eightfold, and the integration of agentic and generative AI technologies. Karnati contrasts deterministic microservices with the more flexible context-aware functionality provided by WIBEY. The discussion offers valuable insights into AI’s potential to revolutionize developer workflows and streamline business processes, according to Karnati and theCUBE analysts.
In this conversation from theCUBE + NYSE Wired: AI Factories – Data Centers of the Future, Walmart’s Sravana Karnati, EVP of Global Technology Platforms, joins theCUBE’s Dave Vellante to unpack how Walmart is operationalizing agentic AI at enterprise scale. Karnati explains how Walmart’s “triplet” hybrid architecture – spanning multiple U.S. regions across public clouds and Walmart-owned data centers – stays vendor-agnostic through the Walmart Cloud Native Platform (WCNP), giving developers a consistent build-and-deploy experience. He details the evolution fr...Read more
exploreKeep Exploring
What is Sravana Karnati's role and scope of responsibility at Walmart?add
What is the current status and future direction of the triplet model in the technology infrastructure at Walmart?add
What steps have been taken to ensure consistent tooling for developers across different cloud providers?add
What must occur for the broader enterprise to fully utilize agentic technology?add
What has been the approach to workforce up-skilling in the context of technology and how have employees reacted to it?add
>> Hi, everybody, welcome back to New York Stock Exchange, NYSE Wired and theCUBE's coverage of AI Factories - Data Centers of the Future. And we're really excited because we're now going to look at how people are applying agentic on top of what are data centers of the future as these evolve. So Sravana Karnati is here. Is the Executive Vice President of Global Technology Platforms at Walmart. You've heard of them. Sravana thanks so much for coming into the studio.
Sravana Karnati
>> Good to be here, good to be here.
Dave Vellante
>> So great to see you here. So you were just featured in SiliconANGLE. Paul Gillen wrote a wonderful feature of highlighting some of the innovations that you guys have. But before we get into that, maybe you could explain to the audience your role, what's your scope of responsibility?
Sravana Karnati
>> Yeah, first of all, thank you for having us here. It's awesome to be with you. I lead what we call Global Technology Platforms. So this is one of the horizontal teams. We are responsible for all of the cloud infrastructure, data centers that we own, and all the foundational systems as well. So in a way, we run the technology that runs the rest of technology for Walmart.
Dave Vellante
>> So you helped architect, what we're familiar with as the triplet model, which, if I recall, it was kind of an open stack on-prem, and you had some cloud capabilities as well. I think there was a dual redundancy. Is that still in place, you're building on? How are you evolving that?
Sravana Karnati
>> It is still in place. So triplet is a core part of our strategy when it comes to how we run our data centers in cloud. So we needed to make sure that our infrastructure is as close to our customers as possible. So we have multiple regions spread out across U.S. And in each of those regions, we have three data centers, if you will. So a couple of public clouds and then our own data centers as well.
Dave Vellante
>> Okay. And from a developer standpoint, if I recall, you've essentially written an abstraction layer that makes the experience substantially similar across all those platforms. Is that correct?
Sravana Karnati
>> Right, exactly. So one of the things we were careful about is to make sure that we're vendor-agnostic. So when you have different cloud providers, we want to make sure that our developers have consistent tooling to build and deploy software. And as we did the triplet strategy, we have created what is called WCNP, Walmart Cloud Native Platform, that allows developers to consistently develop and deploy to any of these endpoints, whether it is our own private cloud, one of the public clouds, or our own data center as well.
Dave Vellante
>> And that's the point, you've got a mix of public and private, people call it hybrid, and you want that experience to be essentially identical across all of those so you're not wasting time and context switching. Okay.
Sravana Karnati
>> Exactly.
Dave Vellante
>> So end of 2022, the AI heard around the world comes out, GenAI. You guys obviously do a lot of advanced research, but I presume like everybody, you were a little bit delighted, surprised to see, and then the light bulbs went off. Take us back to that moment, what was it like for you all when you saw that, and how did you imagine applying it then and how have you applied it today?
Sravana Karnati
>> Yeah, so it's great that you're taking me back to that point. Before ChatGPT became available back in November 2022 and then beginning of March or so in 2023 when they announced a bigger release, we were already using the foundational technology that powers ChatGPT, which is transformer. So Google wrote a paper called Attention is Everything, that really is the foundation for all of this advancement. We were already using that internally for our own work, whether it is cleaning up catalog titles and stuff like that or in being able to forecast better for our customer demand. So we were already using that, but once ChatGPT became available, it's like with everybody else, it unleashed a whole series of possibilities internally, and we were able to do a lot of what we were doing already much better, much faster. So it just caught the imagination of everybody in the company, whether it is a technology associate or business associate, everybody joined the bandwagon and immediately started using it.
Dave Vellante
>> And at what point, because we're going to talk about WIBEY, which is your developer-focused super-agent. At what point did that notion of agents, I mean, I don't know if you called them agents at the time, you might've called them digital assistants like some others, but at what point did that enter your mindset? Was it pre-ChatGPT even? I mean, you must have had an inkling of that. And take us through that agentic.
Sravana Karnati
>> Yeah, it's a great question. So Walmart has always been using machine learning and advanced AI technologies, so for forecasting inventory placement and fulfillment and many other... routing and many other things. So there were already a lot of AI techniques built in there, AI problem-solving techniques built into solving a lot of those problems. So the technology and the use cases and possibilities is not new to us. But once this technology became available, what we saw is developers across our ecosystem started embedding those strategies more and more into what they were doing. As an example, we have a portal called developer experience portal. We call it DX internally, dx.walmart.com, it's available internally for all of our developers. And as more and more tools became available, there needed to be an orchestrator that figures out what is the right tool for your context. And so we built all of this context-aware orchestrators that our developers already were using. And then as GenAI became more and more adopted and the industry started talking about agent and agentic world and so forth, we naturally started adopting and using the same terminology as well. Now this progression is pretty natural, because if you look at automation and robotic process automation and being able to do things in way without doing a lot of coding, it's the same kind of evolution that happened with the GenAI and agentic AI as well.
Dave Vellante
>> But you had a series of fragmented tools that your developers were using. How much friction did that cause, and how did you evolve from that to where you are today, and where are you today?
Sravana Karnati
>> Yeah, that is really a big problem that we started to face as there was a lot of enthusiasm, developers saw potential, they started solving all of these problems. They created more and more agents. And as new protocols like MCP, which is a Model Context Protocol came in, you can actually leverage a lot of the rest APIs and other services that you already built for distributed processing, embed that into agentic AI. And so very soon, what happened was we started having all of these tools, and that increased the cognitive load for our developers. And so it became very important for us to see how do I empower the developers, reduce the cognitive load, and give them a tool set that allows them to have access to all of these things that are available but also do it more efficiently? And that's really led to WIBEY. And then like I said, the initial version of that was orchestrators within the DX portal, but we formalized it with our notion of super-agents, not just for developers, but also for customers and associates and sellers and so forth. We then said, "Okay, let's also put out a super-agent for developers as well, so developers benefit from such orchestration."
Dave Vellante
>> And you're using both MCP and A2A, correct?
Sravana Karnati
>> Both.
Dave Vellante
>> So you've got the stateless as well as stateful.
Sravana Karnati
>> Yeah.
Dave Vellante
>> So is WIBEY is the entry point, essentially, for your developers, and then WIBEY orchestrates what the right tool for the right job, is that correct?
Sravana Karnati
>> Right, exactly, exactly. So when you create an agent within the WIBEY framework... Now WIBEY is built on Element. You're familiar with Element or...
Dave Vellante
>> Yeah, explain Element for me.
Sravana Karnati
>> Element is machine learning and AI platform that allows you to safely build ML models and AI models and safely test them and then deliver in a way that is compliant with our governance policies, privacy security guidelines and so forth. And so the same platform is extended to have a middle layer that allows you to create new agents, register them, register the capabilities, and then WIBEY as an orchestrator has access to all of that information and dynamically can present the right agent for you based on what you're trying to do.
Dave Vellante
>> Because you were coming from what was essentially a set of bespoke tools and the data model was siloed, how did you deal with bringing all that data together, harmonizing it so that the super-agent could confidently act?
Sravana Karnati
>> Yeah, so it really depends on how do you register the agent and agent's capabilities, so the ability for us to then for WIBEY to understand those attributes, and then translate natural language syntax and map that to attributes that agents have, and then pick up the right agent, and then surface that in the right context. So that's really what it is. And in the process, it also knows the context as well. It is not just invoking an agent but also pass the context as well. So a lot of strategies that we've already used in microservices we were able to leverage for agentic AI as well, and that's one of the powers of building your own platforms and having all of this in... You talked about bespoke architecture, because we have that, we were able to very quickly extend that for agentic use cases.
Dave Vellante
>> Interesting. Okay, but I think of microservices, Sravana, as hard-coded. I mean, WIBEY is it using probabilistic AI, is that right?
Sravana Karnati
>> That is right.
Dave Vellante
>> And could you help square that circle for me?
Sravana Karnati
>> Yeah, so you're absolutely right. So microservices, you're much more intentional, it's a deterministic way of solving a problem, you know exactly what you want, but you still need to know where a particular service is running. But even that is deterministic because at the time of deploying it, you know exactly what the end point is, where it is running. At the discovery time, someone can say, "I want this API. Tell me where this API is," and then go get the end point and run it. In the case of agentic AI, as you pointed out, it is much more probabilistic. And that's where based on the context WIBEY basically translates that natural language type of processing translates that into picking the right agent or right MCP server in the back-end and invoke that.
Dave Vellante
>> Okay. What has been the impact on developers, developer productivity? What's the business impact that you've seen?
Sravana Karnati
>> Numerous. So the use cases are every day we discover some new use cases, and that is happening. It's also a reason why it is important to have a super-agent to orchestrate all of these use cases, all these agents. I'll give you just a few examples. One is we create an agent to understand the gaps in accessibility. So when we put out our commerce site and create tools for our associates and so forth, we want to make sure that those tools are accessibility-compliant. So there is a Website Content Accessibility Guideline that's available, we want to be compliant to that. When there are gaps, we've created an agent that actually looks at the compliance guidelines, looks at your code and can identify what gaps there are. And another agent comes right behind it, takes those gaps, and automatically fixes the code, automatically tests it and deploys to production, not quite deploys to production, but create PRs, pull requests, that a developer can investigate and then accept for production. So here is an agent that actually automatically figures out the gaps in code and fixes those bugs and is able to create PRs. Now this has improved the velocity of fixing these kind of bugs by eight fold. So that is developer-
Dave Vellante
>> 8x?
Sravana Karnati
>> 8x. That is developer empowerment because there's hardly any touch in this particular process. Now not all gaps, accessibility gaps are discoverable by an agent or any other automated means. About 60% of those bugs can be automatically found, and about 95% of them can be automatically fixed. So there is almost no touch here for this kind of processing.
Dave Vellante
>> How much of that capability is relatively new as GenAI advances? And I think about Claude having been very strong with developers and coding. GPT-5 was meant to make improvements. I talked to developers, they say, "Wow, they're fixing my GPT-4 code." How fast is that transformer accelerating to create that type of productivity, and what does it look like in the future?
Sravana Karnati
>> It is accelerating faster than anything I've seen so far, and I've been in the industry for several years now. I don't want to give out exact number of years, but it is accelerating really, really fast. And it is also possible that what we develop today could be automated away. So it is important to keep track of the developments in this industry and adapt along with that. So it's everyday a game for us.
Dave Vellante
>> From the standpoint of guardrails that you had to put in and governance, did you have to build, I'll call it an agent control framework? Was that something that you built? Did you get that from MCP or A2A or off-the-shelf technology, or is it a combination?
Sravana Karnati
>> Yeah, the core platform is built by us. We obviously adopt MCP protocol and A2A protocol, and Google just announced AP2 protocol as well for payments. We will adopt the protocols available there because it makes our life that much easier, and interoperability becomes easier. But we have the core platform. Like I said, WIBEY is completely built in-house. It is leveraging the LLMs available and all the agentic capabilities those LLMs have already. We built all layers on top of that, because we want the agents, super-agents to understand our ecosystem, the way we create our configurations, the way we deploy to production, the way we wired our environment, all of that, our agents need to understand, so that's where WIBEY comes into a picture. And then Element, as you talked about, is already a platform that allows you to create machine learning and AI models. We just extended that for WIBEY use case, agentic use case.
Dave Vellante
>> It's my impression, Sravana, that the efficacy for coding is much greater than mainstream workflows. So my question is, where do you see this going in terms of the more mainstream balance of those workflows? What has to happen before the broader enterprise is able to take advantage of agentic technology?
Sravana Karnati
>> So you're right. So the potential for transforming software development has been pretty immense, and so developers welcome that because a lot of their tasks are simplified now. You can actually go upstream and then create product requirements as well. So we created an agent called PM Assist that helps you with that as well, which is actually driving, for those product managers that adopted it, up to 60% productivity improvement. So it is improving the productivity for product managers, improving productivity for developers as well. Over time, it is going to get better and better. It is not just during the development phase, but also after you deploy production or during deployment of production. For example, when I create a code and then I want to push it out to production, there is what is called a CI/CD pipeline, Continuous Integration Deployment pipeline. And if there are issues with that, we've created a pipeline visualizer that automatically figures out what problems there are and also suggests, based on GenAI, how you might go about fixing it. And in many cases, it's just an inspection and the clicking a button and have the agent automatically fix that issue, and off you go. So that's making developers' life much, much simpler, much more enjoyable so they can focus their attention on tougher problems. And as this continues, this go into production as well, when changes are made in production, you need high degree of confidence that those changes are not going to create a unknown side effect. And WIBEY can actually find that as well because we have SRE agents as part of WIBEY umbrella. And when it comes to the rest of ecosystem, business ecosystem, their work process as well is transforming through the tools we are making available. For developers, it is WIBEY. For associates, there's another tool that we're building, super-agent, that would help their associates as well in guiding their inventory workflows, in guiding how they should stock, for example, on the shelves and so forth.
Dave Vellante
>> Well, I hope we can have you back to talk about that development, but last question for you is, I hear the narrative all the time, "AI's going to take away jobs. People shouldn't take computer science anymore." I say, "I don't know, the fundamentals still matter." How did you deal with the workforce up-skilling? I wonder if you could address that, and what the initial reaction was, and how you addressed getting people to lean in.
Sravana Karnati
>> Within the developer ecosystem, within technology, we've not had any challenge in adopting technology. Developers are always waiting for another tool that makes their jobs simpler because there's always a lot more work than we have people. So that has not been a challenge. We do have, however, a Walmart Academy that helps train our associates in all different domains, and we also have Global Tech Academy that provides targeted training as well, including GenAI and so forth. We also have established a high-powered fellow ecosystem within our global technology, and those fellows are individual contributors, very deep, highly experienced fellows who are working with junior engineers to up-skill them as well. So all of these potential exist, but I've not actually had any challenge getting these technologies adopted by engineers. It's in fact the opposite in that people are running really, really fast, and we want to reduce the chaos and reduce the cognitive overload by creating super-agent so that they can find everything that they need in one umbrella and become even more efficient.
Dave Vellante
>> Walmart's been an amazing collaborator with theCUBE. Really appreciate you coming on. Thank you so much for you time.
Sravana Karnati
>> Thank you so much, thanks for having me.
Dave Vellante
>> Great to have you. And thank you for watching. Keep it right there, John Furrier will be back with our next guest. You're watching Data Centers of the Future - AI Factories on theCUBE and NYSE Wired. I'm Dave Vellante. We'll be right back right after this short break.
Dave Vellante