MemVerge Fireside Chat with Steve Yatko of Oktay

Yeah, we do have, uh, other products that integrate Checkpointing technology where we have dozens of customers. Okay. But for the, uh, memory machine, AI is starting in February.

Okay. And, um, um, because we don't have a live customer yet, uh, I invited, uh, an industry expert, Steve Yako here to join me now. I think we were a little slower than expected.

So it's gonna be, uh, we have about seven minutes, so we'll have a quick seven minute, uh, conversation. Uh, um, I think Steve is on Zoom here. Um, maybe to start, Steve, if you introduce yourself, uh, your current role as well as your past experiences.

Yeah, thanks Charles. And, uh, you know, pleasure to be here, uh, with everybody. Um, I'm sorry it'll be a little quick, but that, that's okay.

We'll, we'll cut to the chase. Um, yeah. My name is Steve Yako.

Um, I've got 30 years of, uh, experience, a little bit more than 30 years experience, uh, uh, heavily focusing on, uh, wall Street. I was for 15 years of that. I've spent my life at Credit Suisse, where half of that was a CTO investment bank and building all of their low latency trading systems, a lot of the enterprise HPC.

So it's nice to see it all come back and, and in a form of like a renaissance here to, to be a partner in, uh, the, in the AI world. Um, but I, I was in charge of all the low latency trading and analytics at a global level. Uh, then for my second half of my career there, I ran innovation for all the credit sus group.

Uh, so I owned the visions for applications and infrastructure, uh, and I owned all the labs, uh, to put it all together and test it out and work with all the emerging technology from the startups to the large, uh, it, uh, suppliers. Um, in my last 15 years, I've been very focused on, uh, a very boutique advisory, uh, practice where I continued to advise some of the largest financial institutions in the planet. Uh, as you would suspect, many of them are on this journey, uh, with ai.

Um, uh, it's, you know, been quite an interesting journey so far and certainly have a long way to go and happy to have some of that part of that conversation today in this chat. Um, but I also work in other industries such as, uh, you know, the third largest transportation agency in the us, uh, and the largest, uh, architecture and design firm, uh, in the us. Uh, and very much all of it's focused on, uh, very advanced IOT to field level AI on the edge to core data center, um, training and, uh, inferencing, uh, specializing in, in most cases around how do they make good use of generative ai, uh, and how do they deliver truly enterprise class systems of intelligence using generative ai.

So I'll leave that introduction there, Charles. 'cause I know we're very limited and I want you to have all the time you need to kind of ask questions. Yeah, sure.

So as you advise some of the largest enterprises, uh, regarding their AI initiatives, what are some of the challenges you notice as they stand up their AI infrastructure? Yeah. Well, I mean, it really depends on where they are in their journey, right?

It's, uh, you know, while AI's been around for, for a little while, uh, on the street, um, you know, it's been more of that traditional data scientist ai. Um, but really, uh, the generative AI has taken, you know, pretty much the industry by surprise a couple years ago, um, very unprepared for kind of this journey that they were going to be going on. And I think I heard you say earlier, you know, most of this is being driven by, you know, CEOs to heads technology and partnership.

Um, and you know, I think that's very exciting for all of us. You know, this is very reminiscent of how the world of trading systems got built. We, we moved from a very legacy of very manual environments.

Um, and it was only through the sponsorship of the, uh, business working in true partnership with technology, uh, to really build something as disruptive as trading systems were. And I feel that this is a whole other world of, uh, and it's why I called it a renaissance. I, I really believe it is.

It's, um, it's a very exciting phase for Wall Street to embark on this journey. So it's everything from what is generative ai, you know, the basic level stuff to, you know, now, you know, what are the right use cases, uh, what are the ones most, uh, effective for the business, uh, in terms of driving real, uh, p and l and revenue. Yeah.

Uh, others in terms of where it creates better, uh, customer stickiness and attraction. And then some of the more, you know, complex things around, you know, data that's very confidential and private, you know, such as HR and, and planning career pathways, uh, all the way to the most simple things like, you know, Hey, how do you, you get better, uh, development productivity, or how do you get better documentation, um, available? So it's a wide range.

And, and I think you've highlighted one of the most important things in my view that's strategic around the, the, the journey where it will take them in, in similar challenge no matter what part of the world they're trying to solve in use case is truly the manageability of the application workload. Mm-hmm. And, and the utilization of that fabric.

And for me, also, it's very reminiscent. We spun out dynamic ops from Credit Suisse and the time we built our own virtual environment management framework that ultimately spun out and VMware acquired, and it became the core of, you know, their, uh, cloud engine vCenter, our lead developer became the CTO cloud for VMware, uh, with very successful, you know, the whole point of us doing that was we saw great value in the virtualization technology. You talk about the mig, you know, it's, it's great virtualization technology that is there.

Hey, hey, Steve. Yes. Um, yeah, most of the financial services companies I've worked with in the past, uh, have been fairly free with their money, with respect to infrastructure as far as they could, you know, they would buy whatever they could to get to the advantage that they need for, you know, low latency trading or whatever.

I mean, the effect of building data centers closer to where, where the trading happens and doing fiber optics between them and stuff like that. Why do you think something like member has a place in the financial services? It is really easy, you know, based on everything you just heard the team talk to you about.

It's, it's a super impressive, um, technology platform they're building. And it's right in that sweet spot of what I was describing we did in virtualization. I mean, it's one thing to have virtualization on the hardware, but the value is to extract that and allow applications to seamlessly leverage it to mobilize development and testing, to have the ability to rent and share and lease and put things on ice when you're not, you know, needing them at that moment.

And most importantly, to share showcase, you know, how to actually drive that utilization up very dynamically without manual intervention. And, you know, just people build environments that were virtual and they would leave the firm and no one knew about this virtualization sprawl. So the manageability, uh, of, of job scheduling, the manageability on the orchestration, and then all the economics that can be driven from those insights and, and that mobilization of resources is hypercritical.

So yeah, they used to, you know, certainly they have budgets to go buy what they need to drive revenue, but that's not the old world of, Hey, we're just gonna buy everything and, and, you know, just throw money at the problem. It doesn't work that way anymore on Wall Street. Yeah.

So they're very practical. This is a solution that member has that's highly differentiated. It allows firms to, in the future build their own internal spot markets.

You know, where departments have, you know, where development and test environments that are shared, or just the department has productivity, you know, excuse me, production arrays that have been, you know, put in place for, you know, spike workloads, but they have no way to give that back up to anyone else who might wanna share it elsewhere in the firm. So their ability to, you know, checkpoint and restore, I mean, literally it, it changes the game on your, your firm's ability to just drive up broad usage and shareability and prioritization so that if I do loan out for a dev test environment on my production environment on, I need that resource, I can use the member technology to checkpoint that down into their storage and then resume it on a different node or just maybe there's no nodes available and it gets paused until, you know, that comes up. So that's very differentiating.

Uh, I've, you know, been looking in this space in general for a long time 'cause it is a problem that firms are needing to solve for. They don't, they know from the history of what didn't work and why it's valuable run AI getting, it doesn't even have a portion of capabilities that the member team has and where they're going. But the point was, it's now locked up as an Nvidia only people need multi-cloud, they need hybrid cloud, they need multi GPUs for inferencing versus training.

You know, there's so much more innovation that's gonna come that come out of this small language models are gonna be there, you know, hyper tuned models that are very domain specific, like they're gonna run differently on different types of processors and members will be prepared to support any organization that basically needs hybrid environments, uh, across the things I just described, and give you complete visibility and transparency and versus some opaque or manual environment. So it's very exciting, uh, to see what they're building and the value is going to be hypercritical for no matter where the firm is on the journey, they're gonna land in the same place. Thank you very much, Steve.

I think we're a little over our time. Uh, so, uh, yeah. So maybe, uh, uh, yeah, I think Steve said it more than we could.

Uh, this is not only a cost saving solution, but is a productivity booster and it helps bridge the silos between the departments and, uh, and really allow, uh, the full utilization of the resources and the, uh, unlock all the productivity from across all the teams, firms, Firms will not be able to afford inferencing right on the data without a member capability. I mean, that's really what this will come down to. This is the only way they'll be, be able to afford to do it at a, you know, national or global level, um, on their application of ai.

So thanks for having me. Um, I, I'm sorry we were a little short, but I, i, I really enjoyed the conversation and, uh, listening to everything you guys are building, uh, Charles, and, uh, congratulations to your team and thank you again to everybody for having me. Thank you, Steve.

Bye-bye. All right. So just to wrap up, uh, thank you for spending the afternoon with us.

You know, as we went over, uh, you know, we are introducing sort of a sandwich middle layer that is the, uh, what we call AI infra automation software layer that serves as a good bridge between the workloads and the new GP centric infrastructure and the memory machine AI will automate the deployment of workloads and we are starting this pioneer program in February. Contact us to sign up and, uh, uh, we, we believe 2025 will be a milestone year for enterprise ai. And, uh, we want to do what we can to help the enterprises, uh, to give AI wings.

Thank you.