Our mission is to power the next wave of healthcare breakthroughs, platform that lets innovators focus on curing diseases and not managing the servers, not managing all the architecture and not thinking about the power supplies and so on. We really, really take care of the hardware so that people don't really have to worry about it. They can concentrate on the science and that's what this team did. So choosing the right CRISPR system, designing specific guides for RNAs, picking delivery methods, generating protocols, and even analyzing outcomes. That's a workflow that normally takes months, if not years. What they've done is automate that. It takes a long time to release a drug to market. If we can compress the discovery timelines with the power of AI, that is going to be the game-changer.

Good morning, everyone. I'm so excited for today's podcast because we have Dr. Ilya Burkov from Nebius coming to join us. He is the global head for healthcare and life sciences at Navias and they're really pushing the more important life sciences as well. So Nebius is essentially a global company, which is already on NASDAQ, 8 billion in revenue, massive company. Can you tell us about yourself and what your journey has been all the way from microscopy or being in the lab to becoming basically a server in Nebius?

Yeah, so my name is Dr. Ilya Berkov. I'm the global head of healthcare and life sciences. I lead the healthcare and life science vertical at Nebius. I'm based in the UK. I've been leading the vertical for about one and a half years now. I joined with over 15 years of experience within healthcare, life science and business. I came in as a subject matter expert, basically speaking the language of the scientists and the clinicians together with the business and the hardware side of things. Before Nebius, I spent about three years at AWS Cloud Services, which was a great experience in generic cloud. And they were pioneers for a lot of what the industry is doing these days. And I'm very grateful for my time there. But here we're building something far more fast moving and fast paced and very, very exciting.

Early in my career, I worked in the NHS, the National Healthcare System in the UK, worked in the orthopaedic side of things, did some research as well on biomarker identification for early disease onsets. And before that, I worked in regenerative medicine and also dabbled in some biomedical engineering as well. And what I love the most about my work and what I'm doing today is bridging the world.

of biology, the data and the cutting edge AI infrastructure that's, you know, helping biotech and healthcare teams really accelerate the discovery that they're doing, streamline the R&D processes and bring new therapies to patients much faster. So instead of working on a community level, I'm working on a global scale where I'm making an impact on the healthcare system and on people's lives at AI level speeds. So very, very fast. And yeah, very excited to be here. I'll tell you a little bit about Nebius. Nebius is an AI powered cloud platform, which is vertical agnostic. So we work across robotics, fintech, media and entertainment, and of course, live sciences and healthcare, which is my bread and butter and what I'm really enjoying. But we are essentially helping with drug discovery, genomics and health tech, amongst other things. As a company, uses NVIDIA GPUs for a lot faster diagnostics, a lot faster research and therapy development. The technology that they've been able to create is mind boggling because it accelerates so many things that we do. And Nebius makes it easy for organizations to basically build tune and deploy AI models that they're working on to the next level, making it as easy as possible to do with as little hands-on as necessary because they don't need to manage the hardware themselves. Everything is done in the cloud very quickly, very efficiently and safe. So, Nebius itself is headquartered in Amsterdam. It's a NASDAQ listed company. Again, it is startup, but quite a unique position because we have a global footprint. We are NASDAQ listed. We're building one of the world's leading AI-centric cloud platforms. We have data centers around Europe. We have data centers in the US, and we're building constantly. And this year is going to be a massive year for us in terms of growth and development. For me personally, our mission is to power the next wave of healthcare breakthroughs with top of the range Nvidia GPUs and a platform that is cloud native. A platform that lets innovators focus on curing diseases and not managing the servers, not managing all the architecture and not thinking about the power supplies and so on.

So Harvey, what are your thoughts with companies like Nadeus who are finally concentrating on bringing that technology into our hands so we can actually do this to accelerate a significant amount of improvement in drug discovery? So your thoughts on, because you read this a lot more than I do, so.

Yeah, no, this is huge. This is why they got the Nobel Prize for this particular product. Because think of the impact that it did on humanity. Think how if you are now able to know what we don't know, and my favorite phrase is work backwards and know what that molecule that we're trying to pinpoint is, and then reverse engineer it to the point where we actually now create that key lock effect. For example, if I want to create a certain effect, those drugs are literally sending messages and those proteins are being turned on or off. And that's literally how this drug works. Here's why I wanna take this. If we can use this technology and you just saw those numbers, two to three billion and cut the costs down to let's just say a hundred million and 200 million. And I know that sounds crazy, but what happens is two things. One, that opens the door to personalized medicine. That opens the ability for all of us listening. How can that medication be different for me?

Everyone listening, if we pretend we all had hypertension and we all were on the same drug, not all of us need the same dose. But if the manufacturer only makes certain types of dose, we have to force our bodies into those doses. But if I have this type of medication where I can design it specifically for you and I can create the medicine for you, now I'm creating personalized medicine and this technology will get us there.

Beautifully well said. again, I just want to emphasize this, that of course you can have AWS and you have other things and they are general purpose clouds and they're fascinating products. Don't get me wrong, we use it all the time. Companies like Nibia is when they actually have in-house experts, they can help people to actually go ahead and be able to also teach and learn because you have to understand that as a physician, it took me a while to understand technology. And there's just so many facets, right? So Dr. Berkov, how can Navias help me with CRISPR-GPD, alpha-14-14 and everything, because we know that generic clouds are awesome, don't get me wrong, we are happy with all of those services. But being you as a person teaching, helping us calibrate the technology part while we concentrate on the Medicine for Art would be amazing. So go ahead.

Yeah, sure. So in general, CRISPR-GPT is a fantastic example of what happens when specialized AI meets real scientific need. And it shows the difference between good infrastructure and not. So the infrastructure makes a big difference behind the scenes. And what the Stanford team did together with Princeton and a few other folks, It was that they built something that was not just another model, but a multi-agent system. It's a system that guides researchers through every step of a gene editing experiment. And gene editing is not easy. So choosing the right CRISPR system, designing specific guides for RNAs, picking delivery methods generating protocols and even analyzing outcomes makes it quite unique. And that's a workflow that normally takes months, if not years of deep expertise and technology to master. What they've done is automate that. Essentially giving that technology and that knowledge into the hands of folks that don't need to have a postdoc or a PhD in this. They can have this system and this agentic approach to something very, very complicated in, you know, a master's degree level, dare I say it, even a bachelor's student. And what's really striking is the real-world validation. So junior researchers with no prior gene editing experience have used CRISPR-GPT to design and execute very complex experiments. And when you're looking at knocking out multiple genes in a human cancer cell, it's quite difficult to do. But with this tool, they've achieved around 80 % editing efficiency in their first attempt. And this is just the starting point. With other experiments that they've been writing about and this is published in Nature Biotech. You can see that they're hitting over 80, sorry, 90 % success as well with other experiments. These are results you'd normally only expect from very, very seasoned experts. And so what made that possible? It's AI that is designed with reasoning. It is a system modulator and it is a modular approach with multi-agentic architecture. You need compute that's iterated. So it is, it has taken a long time to be able to provide this infrastructure, to be able to rapidly prototype and fine tune very specialized models. And, you know, at Nebius, this is what we do. We really, really take care of the hardware so that in the background people don't really have to worry about it. They can concentrate on the science and that's what this team did.

Dr. Birkov, we talked about Drift for GPT. We're going to do slicing, understand, rather than moving from wet lab to dry lab and then actually focusing on exactly our efforts to bring the wet lab, basically used for therapy, including biologics and everything. Just help me understand the landscape of drug discovery and how Nebius and you are helping companies in terms of drug development. Do you have something for that or you using it in the planning?

Drug discovery is extremely complicated process and there's no one system that does everything. What we do is we enable the infrastructure and we provide the infrastructure for that to be done. We work with a number of groups looking at drug discovery, both on a startup level and beyond. Looking at something that is a complicated, it takes a long time to release a drug to market. If we can condense that period into something that's a bit more manageable, if we can compress the discovery timelines by combining a lot of these high throughput screenings with the power of AI, I think that is going to be the game changer drug discovery into a real-time data optimized process. So AI models are able to analyze the results as they come in and essentially guide the next set of experiments, which can potentially dramatically speed up iteration cycles. And essentially, the end goal is to compress the timelines from over decades to perhaps years, or maybe even in some cases, a few months. Because the data quality is higher, the candidates are more likely to succeed clinically. And we're not just moving faster, we're making much better decisions earlier.

So amazing that you just said that. And then Harvey, why don't you expand on what he said, agentic AI in CRISPR and in drug development, because you are a master of that for people out there, obviously it's complicated process, but if you throw in an agentic part and imagine that in the, from going from A to Z to a drug that you could create agents that would do that segment, do the research, do that, present, then you have an orchestrator and then you're adding this technology. And to what his point, he said, this is doing it better and more accurate.

And so by that statement alone, even if they were able to decrease the error rate by a certain amount of percentage, then that multiplies because now that end of product, Jelaine showed that, what was that PowerPoint that said 90 % of drugs end up in failure. And if you can lower that percentage, then gosh, think of all the savings. One of the things that we didn't catch offline, because we had some technical difficulties is I truly believe this is the road to personalized medicine.

Now, so I just want to add to Harvey's point and Dr. Burkovs' point is again, what we are moving towards is highly specialized models that are done recursively in an agentic way. And when you have this new system in which there's reasoning capabilities and recursiveness that goes through that, creating solutions is extremely, extremely important that the hardware software works together. Because you do not produce quality results. Because we do that all the time, these experiments with different providers. And that's again why we are excited about putting this show up, is just making sure that people understand. I want you to sort of tell us more about Navias, not just the help. So that, you know, there are probably people who are founders who's looking at Navias for different reasons, including maybe robotics or something.

And then one last thing, if you were to see three years from now, because five years is no longer possible, nobody can make a decision, but I'm just saying, in three years from now, specifically in health sciences, what is your prediction?

Yeah, well, fantastic. So Nebius in general is growing at lightning speed or as I always like to say, at AI speed. And what we are doing now and where we're going next is being able to provide GPUs in a democratized way where we are able to give the latest and greatest chips to even the smallest chip companies and startups that are working with us. We're not picking out very large enterprisey companies only to provide the latest and greatest, which you often see with hyperscalers. We have this approach where whatever the workload that you're doing, you can get it on Nebius. Whatever storage you need, whatever requirements you need, whatever orchestration tools, you're going to be taken care of. And there's a level of handholding that I've not seen in my previous experience of working in hyperscalers, which is very white glove in its approach. It can be as hands-on or hands-off as you want, but often we are brought in at the beginning to help, you know, virtualize everything, prepare everything, and then let the experiments run that the scientists do their work. It's very difficult to predict, but it's exciting to see how many different points are going to be breakthroughs. And I think that the defining breakthrough, I guess, if I have to choose one, would be AI systems that can reliably predict patient outcomes before treatment begins. So not just at a population level, but for individual patients. And that's similar to what Harvey was saying. That means digital twins that combine genomics, imaging, real-world evidence, and longitudinal clinical data to simulate how a therapy will work before it's even given. And what makes it possible isn't a single algorithm, it's the infrastructure. So in a nutshell, I think that the breakthroughs won't be smarter AI, it'll be more predictive. patient specific medicine at a larger scale that's enabled by cloud platforms, delivering much more raw compute, the security than the compliance that's required and trusted by a single or multiple integrated systems. That's how I would position it.