theCUBE + NYSE Wired: Zero Trust Cyber Series | Ahmad El-Dardiry, President & COO - AI & HPC

Clips
More from theCUBE + NYSE Wired: Zero Trust Cyber Series

Ahmad El-Dardiry

President & COO - AI & HPC

SourceCode

play_circle_outline Introduction to SourceCode's Gryf: AI supercomputer in a TSA-approved suitcase.

play_circle_outline Exploring Edge Networking and Packaged Innovations: Unlocking Industry Advancements

play_circle_outline Maximizing AI Potential: SourceCode's Approach to Building and Integrating Systems with Data Sovereignty and Governance

play_circle_outline Enabling Enterprise Gen AI with Orchestrated Pilots and White Glove Service from SourceCode

play_circle_outline Configurability and adaptability of technology

play_circle_outline The Future of AI in Enterprises: Importance of Collaboration, Innovation, and Risk-Taking towards 2025

Info
Transcript

Ahmad El-Dardiry, President & COO - AI & HPC

Ahmad El-Dardiry

President & COO - AI & HPC SourceCode

Welcome back to theCUBE at the NYSE for three days of Media Week, Cyber Week, and AI Innovations. SourceCode is partnering with the CUBE on AI endeavors, including Gryf, the world's first AI-enabled supercomputer in a TSA-approved suitcase. Gryf is configurable and self-contained, perfect for edge computing. The industry is changing rapidly, and companies need to think about multi-solution deployments. SourceCode partners with various vendors to provide customers with the best infrastructure solutions. They are excited about generative AI and the impact it wi... Read more

explore Keep Exploring

What is the name of the supercomputer with AI-enabled capabilities that can fit in a TSA-approved suitcase? add

What is inside the TSA-approved case being discussed in the conversation? add

What is SourceCode's approach to working with customers in the AI space and how do they partner with Nvidia and other technologies to provide center of excellence solutions for enterprises to experiment with AI on-premises? add

What approach has been taken to help companies stand up their proving ground, proof of concept, and center of excellences for AI innovation with good utilization and orchestration? add

What are the key aspects of our approach to partnering in the rapidly changing industry? add

What are you most excited about in terms of investments and innovation in the next couple of years, specifically in relation to generative AI and its potential impact on businesses and their bottom line? add

bolt Powered by CUBE AI

Ahmad El-Dardiry, President & COO - AI & HPC

search

>> Hello. Welcome back to theCUBE, here at the NYSE. I'm John Furrier, your host. Dave Vellante was here earlier. And this week here, for three days of Media Week, Cyber Week, AI Innovations, and we're here in the East Coast CUBE office and studio, the super studio. They got the show floor. Of course, we've got the Silicon Valley connecting Wall Street with Silicon Valley and bringing all the best content to the community as part of the CUBE NYSE Wired Network, and now partnership with SourceCode. The CUBE is partnering with SourceCode on a variety of AI endeavors from AI labs and factories. We're going to be building with benchmarks, use cases from customers. A lot of great stories. We're going down in the weeds and getting all the data, as our role to get that and share it with you. And Ahmad El-Dardiry is here, president and COO of SourceCode running the AI and HPC business. Great to have you on theCUBE and your special guest that you brought along with you, Ahmad. That's quite the guest we got here. Explain who's with us here and then we'll get into the SourceCode.

Ahmad El-Dardiry

>> Yeah, John, it's a pleasure to be here in NYSE and be part of the NYSE and theCUBE Wired community and the AI innovators. This is a very special guest. I'm pretty sure this is the first time it's on camera. This is Gryf, and Gryf is at the world's first supercomputer with AI-enabled for the edge in a TSA-approved suitcase. Our team has actually tested this. You could go through the TSA and put it in an overhead luggage. It's something that we've been working on for quite some time, and it's built by us and designed in collaboration with GigaIO, with their fabric technology. It really came out of a collaboration with US governments a couple of years ago when we were thinking through a lot of the problems on data transport, mission applications that's out in the field without connectivity, and how you go do AI workloads out in different missions. But the application here is endless, and yeah.

>> Of course, the government would want a tactical edge, a battlefield scenario or any scenario. But I just want to ask, if you don't mind, so this is approved by the TSA, you said?

Ahmad El-Dardiry

>> This is in a TSA-approved case.

>> Case. So packaging innovation.

Ahmad El-Dardiry

>> Absolutely.

>> What's in it?

Ahmad El-Dardiry

>> So in it, it's actually quite configurable and it's a 19-inch rack-mount case. It has multiple sleds. It has an NVIDIA GPU sleds and a couple of compute sleds, as well as an SSD ruler form factor sled, a networking sled, and of course, power. It's self-contained. You could have it connected to your core infrastructure, or you could have it out unconnected and still be able to do intelligence work, signal, data collection and assessment on the edge. But also, you could link up to five of them. So you could truly have up to 30 different nodes talking to each other and working together out in a field, or in some commercial applications that we're seeing really interesting ideas coming from customers in media filming on location, sports analytics, oil and gas.

>> Talk about edge of the network capability. Not to digress, not to show my age and all, but when you were carrying that into the studio here at the NYSE, it's got a little handle in the front too. It's got the pull-up like I got a suitcase, it reminded me of the late '80s, early '90s when Compaq came out with, actually it was the '80s, when Compaq came out with the first Luggable. They called it a Luggable laptop. And if you remember back in the day, it was basically a laptop, they called it.

Ahmad El-Dardiry

>> That's right.

>> And it was about that size. So what I see now in my mind is so clear is that that's going to accelerate on the packaging side. It's going to get smaller, it's going to get faster. It just really shows the innovation of what we've come through in the industry. Essentially, that's a data center.

Ahmad El-Dardiry

>> It's a data center on wheels.

>> It's a data center on wheels, and that's phenomenal. Love it. Thanks for bringing that on. We can continue to talk more about it, but I want to get into your business because we have been collaborating with the NYSC, with the Wired Network, with Brian Baumann and Anand and the team here. We see all the technologies coming through here, thanks to Anand and the AI Excellence Center as part of ICE and really amazing organization. They have a high bar for technology, and I got to say, I've learned a lot and shared a lot with them, working on a collaboration with AI, getting AI factories up, AI labs and getting things like this into the hands of startups and enterprises, and you guys have been doing this. And I want to have you explain, because I've learned through working with you and your team, the capabilities, it's state of the art. I want you to take a minute to share SourceCode as a company, some of the things you're doing, why you exist, and some of the tactical things that are relevant right now that I find compelling.

Ahmad El-Dardiry

>> Absolutely. And yeah, the team here at NYSC, Anand, Brian and others, it's been a great discussion and collaboration with them. And also learning. But at SourceCode, we've got a very deep heritage. We've been a company that's been operating for over 30 years. We service customers across 40 countries, and we really work with an engineering-led approach with our customers. So we work with customers of all sizes from government Department of Defense to geospatial, to commercial and enterprises and banking, and health and life sciences, and a lot of university and research. And what we've been focusing on with very deep partnership with Nvidia and other technologies in the AI space is really how to go work with enterprises in their early experimentation of AI with center of excellent solutions that they could really control and own on-prem. And we were all been talking about what does that mean? How do they go encourage experimentation and testing for their use cases and really think through their data sovereignty and data governance?

>> Yeah. One of the things that's come up that I've identified on theCUBE and certainly, we've been collaborating a lot on is this demand. Appetite is off the charts in terms of looking at how people are looking to try to evaluate how to stand up critical infrastructure now in clustered systems, whether it's luggable supercomputers or standing up a data center connected to the cloud to run not only what configurations of that hardware and software stack, but also the configurations of those. There's two dimensions of it. There's two threads pulling there, and then people are throwing benchmarks around. Every keynote I go to, there's benchmarks flying everywhere, and I think this is a huge opportunity to use these center of excellences. The things that you're rolling out as a service I find to be super valuable. And of course, the world needs to know which benchmarks are real, which ones are relevant, because not all benchmarks are created equal. Especially in these emerging times we have acceleration in the marketplace, it reminds me of when the processor of innovation from the 8086, the 286, and then you saw the processor evolution, but there was some standards then. It was a little bit slower. Now, it's not just the thing, it's multiple processors, multiple fabrics. It's a harder nut to crack.

Ahmad El-Dardiry

>> Absolutely.

>> What's your view on this? Because you have a strong opinion on this and I like what you're doing, and talk about the complexity of doing all this.

Ahmad El-Dardiry

>> The complexity is enormous and for someone to stand up a full-stack environment and really thinking through what are the open-source solutions out there, what are the open-source LLMs like Llama or Mistral that they could do in a way that's controlled for their data and especially their critical data. There's thousands of combination and thousands of variables that you really need to think through as you're building that. So what we've done in partnering with customers is really thinking through, in this early journey, how do they take an infrastructure and get the best use out of it, the best utility out of it, as well as thinking through their proof of concept into production? Where are they getting ROI from these use cases? We hear about tons of use cases, a pent-up demand of experimentation. And right now, we're also seeing, in 2024, enterprises were shy in investing, and they've been experimenting sometimes out in hyperscalers who have invested quite a bit. But one of the challenges there is data gravity is real because once you experiment out in clouds and you really need to think through what's your path to production and is the use case, and the data behind it because data is key, that's your superpower, is it appropriate to be in your cloud environment or is it critical to keep it on-prem? So some of the solutions that we're taking in our center of excellence approach with partnering with customers is to really think through the control. How do they build it in a control? How do we help them set up a stack, how do we containerize it? And then how do we, if appropriate, link it to their cloud for production bursts or in building blocks, deploy it for them in their colos or in their own data centers, right?

>> Yeah. When I talked to the JPMC CIO, Lori Beer at Reinvent, fantastic interview if you want to go to the YouTube video on the SiliconANGLE's CUBE, check it out. She said, "We were doing hardcore machine learning." They got $17 billion annual budget, $10 trillion a day, and they're hardcore. Their bar is high.

Ahmad El-Dardiry

>> Yes.

>> They are building systems, but gen AI is coming for them. So you're starting to see them go, "Okay. I'm going to need to have this supercomputing capability." They already have tons of data and they're setting up the gen AI path. Now they're not slacking on gen AI. They're just slowing down production because their resiliency framework is so high bar that they're got to make sure everything's resilient. They need to be 100%. They're not all enterprises, but my point is they're sandboxing. And I call it sandboxing, my term, but generically speaking, trying to create an environment for the innovation to start thriving.

Ahmad El-Dardiry

>> Absolutely.

>> I find that what you're doing is really the benchmark for getting groups to start playing, building for two reasons. Testing workloads, understanding configuration, but also team building.

Ahmad El-Dardiry

>> That's right.

>> Who's going to lead the project? Everyone in the organization wants to lead the gen AI

Ahmad El-Dardiry

>> And it comes from the businesses. It comes from the business units, not from IT. So how do you really orchestrate and encourage pilots, proof of concepts and experimentation without the concerns of runaway costs or the concerns of a -

>> That's what I was going to get to, cost, because now, I don't need to set up, I can go to a SourceCode AI lab with theCUBE or whoever, or build my own and make it the playground.

Ahmad El-Dardiry

>> That's right. Absolutely.

>> And that's where you can get the innovation, and then see how it goes, then invest.

Ahmad El-Dardiry

>> Absolutely. And we work on whether it's a small investment at the beginning with a couple of nodes or at the Racks K level, is it different technologies that you want to experiment with on the inference side? Really partnering also with the best of breed of system vendors, but really standing it up in a customer's environment with a white glove service, helping them set it up, bringing in our data science team to really do the right installation, and sometimes, in the right benchmarks for them.

>> I want you to also share your thoughts on the SourceCode benefits. I want to highlight something and I want you to weigh in on it because one of the things I like about what SourceCode's doing and what you guys have done is you've got a new fresh perspective on what the environment needs to look like. A lot of the things I see out there in the deployments are either old school hacks and benchmarks, or people not qualified, or super qualified in one dimension.

Ahmad El-Dardiry

>> That's right.

>> But they can't scope out across more of a horizontal view of what's needed. They might be specialists in certain areas, but they might miss a lot. So you're seeing a lot of domain expertise that's either siloed and misappropriated or not even figured in.

Ahmad El-Dardiry

>> That's right.

>> This is a key part.

Ahmad El-Dardiry

>> Absolutely.

>> You guys have taken a different approach.

Ahmad El-Dardiry

>> Absolutely, yeah. A couple of months ago, we actually brought in the world's leading banks from around the world here in New York as well as Asia and Europe, health and Life sciences, healthcare. And it was really a round table that was Chatham House Rules, so I'm not mentioning any of these customers.

>> . Love those.

Ahmad El-Dardiry

>> Even universities that are on the medical research side that were present and a couple of large telecoms. And the one common thread is that pent-up demand from actually bottoms-up in the organization to really experiment with the benefits of what gen AI could do from different use cases is really, really strong. The top down requirements are also very, very high. But everybody, there was a common thread that to see an investment both in the infrastructure driving the right utility of that infrastructure, all the way to the use cases and seeing bottom line ROI, enterprises are not seeing that yet, but it's coming.

>> And they have the plan for it.

Ahmad El-Dardiry

>> They have the plan for it-...

>> because the top-down is dictating, "We need to be moving our business model this way, save our costs." Bottoms-up needs all the tools and technologies and platforms that they need.

Ahmad El-Dardiry

>> That's right. That's right.

>> But they're also in discovery mode.

Ahmad El-Dardiry

>> Yeah. That's right. So it was unanimous actually. It was really incredible. We had about 40 chief AI officers, CTOs, technologists, MDs that were in the room. So the approach we've been taking is how do we go help these companies stand up their proving ground, their proof of concept, their center of excellences with really good utilization and orchestration on how they could encourage experimentation within their control. So we have a white glove service. We roll out different AI factories, and we could also work on different underlying technologies as well.

>> Well, I love this Gryf tool, G-R-Y-F.

Ahmad El-Dardiry

>> That's correct.

>> What's that stand for? Is it an acronym? Does it mean anything, or is it more of a-

Ahmad El-Dardiry

>> It's a product. It's a name that we... We had really good partners with GigaIO to really think through the fabric technology that they have and how to drive the connectivity. But this came out of a collaboration or a requirement with the US government actually. It was a big customer within the government that really wanted this AI out in mission, but now, we can't really keep up with the demands of... By the way, it's a couple of months in production.

>> Yeah. How heavy is it? How heavy is it?

Ahmad El-Dardiry

>> You could try to carry it, for sure.

>> while I'm wired in here.

Ahmad El-Dardiry

>> Yeah, you could, absolutely.

>> Like a ballpark. But it's enough to lift and put it in the overhead bin.

Ahmad El-Dardiry

>> Absolutely. You could lift it-

>> It literally could go in the overhead bin.

Ahmad El-Dardiry

>> It's traveled commercially across the US.

>> Well, you're pretty bulky. You're like a triathlete, right?

Ahmad El-Dardiry

>> Yeah. I can .

>> You can actually put it in the bin?

Ahmad El-Dardiry

>> You can absolutely put it in the bin. You could carry it.

>> Okay.

Ahmad El-Dardiry

>> It goes through security approved.

>> That must really turn a lot of heads at TSA.

Ahmad El-Dardiry

>> It really is. And I will tell you the excitement around it, every customer discussion, we walk in with, "Here we think is your use cases." Before we end these meetings, I kid you not, it's incredible the level of excitement. They come up with three or four more applications that we haven't thought of in their environments.

>> What I like about what this shows and some of the things you guys are doing in SourceCode, I don't want to use the word integrator or whatever because that's an older term, but you're putting the solutions together, and we were on a panel last night up in the NYSE Wired Chatham House Rules and I moderated with AMD and Tensorway which is a neoCloud .

Ahmad El-Dardiry

>> Absolutely. We've talked to them as well.

>> You have a neoCloud configuration. You've got on-premise, which needs its own excellent center. You got connected to the cloud. These are not well-known constructs in the mainstream over the past few years. Now even in the past few years, and they're changing. So there's no rubric. There is no playbook for integrating this. You guys are at the cutting edge and this is causing the customers to scratch their head. What does it mean?

Ahmad El-Dardiry

>> Absolutely.

>> They all want to know what does it mean to me because I want to design the best system possible. So in a way, you're like a system builder.

Ahmad El-Dardiry

>> That's right.

>> You guys are building systems.

Ahmad El-Dardiry

>> This is our onboard design. It's manufactured in our ITAR certified security cleared facilities in Boston where our headquarters are. We've got system manufacturing integration and testing in London, Germany, Boston, Seattle. So we're around the world, and we really think through what... We start from the engineering problem backwards, and that could require us to partner with system builders and reintegrate, or partners with networking, with storage or build it ourselves.

>> I'm not going to say it directly, maybe I will, but when I said about the Compaq example, they started in a garage assembling things. Michael Dell started his dorm room.

Ahmad El-Dardiry

>> Absolutely.

>> They built and assembled systems, PCs and then servers.

Ahmad El-Dardiry

>> Absolutely.

>> So we are in another era-

Ahmad El-Dardiry

>> Absolutely....

>> of a wave that's going to look a lot like that last PC server revolution, Wintel, to the Nvidia, AMD, neoCloud , Amazon Web Services, Azure, Google, Oracle, all. And under the hood, HPE, Dell, Lenovo, Broadcom. This is like Silicon diversity meets... Dell would do this.

Ahmad El-Dardiry

>> We partner with Dell.

>> But they're not doing that. They're doing the AI factories.

Ahmad El-Dardiry

>> That's right.

>> HPE's got the big boxes.

Ahmad El-Dardiry

>> Absolutely. I'll tell you about-

>> Lenovo's out there.

Ahmad El-Dardiry

>> I'll actually tell you about another partnership we have. Eviden, which is a big HPC out of Europe, and we are their partner from North America, all the way from Excascale to cloud to liquid-cooled systems that we're working with them on. So we really partner across the board and we firmly believe in the same hypothesis that you're saying. The industry is changing rapidly. Everybody needs to think about multi-solution, multi-deployments. And another thing about this is it's really configurable. So we know that even as a customer acquires 10 and 20 of these, in mission, it changes. The requirements change, the configurability change, the number of compute sleds versus GPU sleds change, the size of storage change.

>> And you're Switzerland, you love Nvidia, you love everybody.

Ahmad El-Dardiry

>> Absolutely. We have a very, very deep partnership with Nvidia. We're one of their premier partners in Europe. We've got a very deep software partnership with them as well. They're definitely on the bleeding edge of innovation. The other thing I'll tell you is while you're out filming on location or collecting data or doing sports analytics, and you're set up there for a period of time, you could pull the storage sled, put it in a backpack, get back to your core network or your core data center, plug it in, download the data, replace it with another storage sled. It's really the flexibility and the -

>> So there's the changing the software picks it up, hot-swappable, all that stuff's built state of the art. You got to love the innovation and if I'm Dell, if I'm HPE and Lenovo-

Ahmad El-Dardiry

>> This extension-

>> I'm rethinking my strategy because I'm saying to myself, if everyone's going to go to build your own cluster-

Ahmad El-Dardiry

>> That's right....

>> am I going to want to own the whole cluster like Nvidia? Maybe I do. Maybe I want to be heterogeneous, but we all know history. Open source, open standards and heterogeneous systems and networks always wins.

Ahmad El-Dardiry

>> Absolutely. Absolutely.

>> Except for the unique use cases.

Ahmad El-Dardiry

>> Collaboration is really important to service all our customers and to get them to get the most out of their infrastructure, the most out of their investments.

>> What are you most excited about? We'll end this segment with your most exciting area. Obviously, the innovation you're doing with the form factor. Building systems, that's great. The innovation around AI competency centers, whether it's labs or other for helping people build their sales. All cool. What else are you super excited about?

Ahmad El-Dardiry

>> Yeah, I'm really excited about the next couple of years, the investments in generative AI and seeing the results and the impact that it could have in businesses and their business processes in their bottom line, whether it's efficiency gains that they get from their workforce or better serving their customers or deploying new products and product lines. That I'm very excited about, and I firmly believe 2025 specifically, this is where enterprises are going to start to take a big risk, calculated risk, invest and scale.

>> Great to have you on theCUBE. Great to see you again.

Ahmad El-Dardiry

>> Yeah, it's an honor to be here.

>> TheCUBE is partnering with SourceCode. We are going to bring benchmarks. We're going to bring data to you. We're going to partner and get all that configuration. We want to set up these factories, these AI systems. We want to put the best workloads on. We want to get those benchmarks. We're tracking all the speeds and feeds. And at the end of this video, you're going to see a demo. We're going to unbox this carry-on-

Ahmad El-Dardiry

>> On the floor....

>> on the show floor if they let you. And be careful. Trump was here. So yeah.

Ahmad El-Dardiry

>> I know. The-

>> Did Secret Service take a look at that?

Ahmad El-Dardiry

>> The president-elect rang the bell this morning, so it was a great honor to be here. We had this dog sniffs on the way in with the Secret Service, and I think Anand, the head of AI here, wants to take it down on the floor, unbox it and-

>> Shout out to Anand at NYC ICE.

Ahmad El-Dardiry

>> And Brian.

>> He runs... Well, Brian gets a shout all the time. He runs the wires . Anand's run will be scaling up. He's doing some really pioneering work around AI Center of Excellence.

Ahmad El-Dardiry

>> Across ICE.

>> Across ICE, but not only that, not only is that notable. What he's really doing is he's setting up a culture shift and he's creating an environment where he's sourcing all the requirements, and he's creating an environment to train and get people their hands on. They're testing. They got so much data and I think he represents and encapsulates-

Ahmad El-Dardiry

>> The bleeding edge,...

>> all enterprises. And if I'm an enterprise, I would set up my own Center of Excellence because why wouldn't you? You've got to get your hands on... And you can't always get the gear.

Ahmad El-Dardiry

>> Absolutely.

>> Okay. You'll see a video right now after this video of the unboxing of Gryf. Thanks for watching.

Ahmad El-Dardiry

>> All right, thank you. Thank you for having us.