theCUBE + NYSE Wired: MedTech Unplugged, the Future of AI in Healthcare & Life Sciences | Omri Yoffe, Vi

Clips
More from theCUBE + NYSE Wired: MedTech Unplugged, the Future of AI in Healthcare & Life Sciences

Omri Yoffe

CEO

Vi

play_circle_outline Non-Disruptive Enterprise AI Orchestration: Turning Healthcare Data into Next‑Best Actions for Trials, Navigation, Supply Chain, Oncology

play_circle_outline Protecting Patient Privacy at Scale: Tokenization and De-identification in Vi’s 1.2 PB/Day AI Pipeline

play_circle_outline Edge deployment with human-in-the-loop guardrails; four deployed engineers for orchestration and compliance

play_circle_outline Governance emphasis: collaborate with government, health systems, and pharma for ethical, compliant AI

play_circle_outline Health Data Firm Nears 200M Patient Tokens, Posts Nine-Figure ARR and Targets $1B Contracts

play_circle_outline Hospitals as AI Factories: Proven ROI, Faster Drug Development, and the Rise of AI-Native CROs

Info
Transcript

Omri Yoffe, Vi

Omri Yoffe

CEO Vi

In this interview from theCUBE + NYSE Wired: MedTech Unplugged, Omri Yoffe, chief executive officer of Vi, joins theCUBE's John Furrier to discuss how enterprise AI is converting massive patient datasets into real-time clinical intelligence. Yoffe details Vi's three-layer approach — a privacy-safe data platform spanning 190 million de-identified patients, purpose-built private models and an edge orchestration layer — that delivers "next best actions" across oncology, clinical trials and supply chain. Processing 1.2 petabytes daily, the system tokenizes patien... Read more

explore Keep Exploring

What is Vi, and how does it help healthcare and life sciences organizations? add

What are the main layers to consider when building AI-driven solutions in healthcare to become a category leader? add

How should organizations approach building and deploying private AI models at the edge—how many models to create, how to scale them, and how to ensure engineering support, compliance, and appropriate human oversight? add

What is your reaction to the use of AI in healthcare, and how should developers address concerns about its life-or-death implications, data privacy, and the need to collaborate with regulators, large health systems, and other established institutions? add

What is Vi's current financial and operational status, including ARR and profitability, investor backing, scale of de-identified patient data, AI/ML product and agent capabilities, and near-term growth/contract value targets? add

Is there a real and growing market for AI‑native clinical trial service providers—including companies leveraging IoT/wearables—to accelerate trials, reduce costs, and create opportunities for clinical‑trial professionals? add

bolt Powered by CUBE AI

Omri Yoffe, Vi

search

John Furrier

>> Welcome back here. I'm John Furrier, host of theCUBE here, theCUBE's NYSE studio here on the East Coast on Wall Street, of course. We are a Palo Alto office and studio in Silicon Valley, connecting Silicon Valley and Wall Street. It's part of our MedTech unplugged series. We talked to the leaders who are making things happen, bringing in technology as technology becomes the market. It is enabling new kinds of capabilities that's changing the world and frankly, bringing in new capabilities that were hard to do or didn't exist, and we've got a great guest here. Omri Yoffe is the CEO of Vi is here. They're doing a lot of interesting things and bringing intelligence into healthcare and really making changes. Omri, great to see you. Thanks for coming in. Appreciate it.

Omri Yoffe

>> Thanks for having me.

John Furrier

>> Okay. So you're the CEO. Looking at the whole landscape of healthcare with AI. Explain what you guys do. I want to jump in because you guys are doing some very innovative things. So narrow down, what specifically are you guys doing? What problem do you solve and who do you solve them for?

Omri Yoffe

>> Sure. So Vi in short is an enterprise AI for health. We are serving the largest healthcare and life science organizations in the world, helping them taking their data, basically translating it into the patient and care team's next best action. And the way we do it is anywhere from clinical trial acceleration, patient care navigation. We help billions of, let's call them next best actions around anywhere from oncology to clinical trials to supply chain. Anything that can make your health journey as a patient better, faster, with a reduced cost of care. And we bring three main assets to the table, very straightforward ones. The first is data. So we are spending over 190 de identified million patients that we basically de identify and tokenize into our system. Second, we productize AI models on the edge level, on the workflow level. If it's a hospital or a clinical trial location or a chronic disease specialty clinic, and we build AI models or productize them in a way that will be deployed on top of existing tech stacks. So you don't need to change any of your tech stack to keep your Epic, your AWS or any types of CRM. And we're building an orchestration layer that sits on top of your existing tech stack. And the third, our business model is very straightforward. We win only if the patient and the enterprise win. So it's not a classic SaaS or a technology play. We underwrite how can we help your health organization. And then we basically create a control group so we can measure the actual efficacy and impact that we're making.

John Furrier

>> So if I get this right, we saw the cloud, we see what AI's doing. Cloud was horizontally scalable compute with SaaS apps.

Omri Yoffe

>> Yep.

John Furrier

>> That moves to distributed computing where you have cloud on prem edge where the scale is a data scale and the apps are domain specific.

Omri Yoffe

>> Absolutely.

John Furrier

>> And the edge piece is becoming key as we just came back from Mobile World Congress or MWC. I just put a report called HyperConverged at the Edge where if you can get intelligence to the edge, it opens up new things that we couldn't fathom in the past. So one, you got to set the table first with some AI. So I want to get first to how you look at the architecture of how you go across all those databases.

Omri Yoffe

>> For sure.

John Furrier

>> How do you get that data plane, a control plane? How's it harmonized? Because there's data out there, but sometimes you just can't get to the right place fast enough. So everything is real time. Next best actions assumes that you actually get the next action. So it's like you got real time response, data availability and speed are all on play. Well, that's hard to do.

Omri Yoffe

>> That's precisely correct. And there's a lot here to unpack, but I think that the three main layers to look at are the following. The first piece, if you want to be effective in healthcare and any other vertical that you want to be a category leader, you need to have a massive amount of first and third party data. Meaning, you need to gather enough info that is not on the edge device. It's on your cloud in a secure privacy safe way and you need technology for it. So without getting into technical details, you need to build both a culture, but also tools to treat privacy safe information, how to tokenize. I'll give a practical example. If I'm heading into a hospital, our system can get my first family names and address and remove it from the rest of my information. So I keep it as an ID within the system. So next time you come and trigger an AI action to my next best action, the system will see it. So that's number one. You need to gather this information.

John Furrier

>> That's a heavy lift just in itself.

Omri Yoffe

>> Correct. And again, just to bring it to life, Vi today computes 1.2 petabytes a day. This is one of the largest computing efforts for first and third party data for behavioral psychographic demographic operational data. So that's number one, that's the first layer. Not on the edge device, on the platform device. The second piece is how do you pick the models, agentic models or others that will drive the best action for your use case. If you're trying to develop a drug, an oncology or kids epilepsy drug, what is the right model to activate and find the right patients and to streamline the operations of the clinical trial in the faster way? So you have here, it's not exactly the edge device. It's more for data science product.

John Furrier

>> It's almost pre-edge because one, you got to get that platform that you just mentioned, which is hard to do, but if you get it done right, get it done. But then having the right models in the position is key. And we're seeing a lot of that specialty model. It's not going to be the big frontier model. You can talk across the network, but this is specific.

Omri Yoffe

>> It's very specific. It's what we call private models. And you don't want 200 of those. You want to cherry-pick four, five, 10 winning use cases so you can scale and make impact. So it's a product with high margins and scale. And the last point that you are very precise on is the edge. At the end of the day, you have an extra 10, 20% that you need to implement and tweak and build compliance and human judgment. And this is where we do most of our work remotely. I can explain why it's helpful, but you also need some four deployed engineers, very capable, knowledgeable people to make sure that the actual AI orchestration layer is doing what it's supposed to do.

John Furrier

>> Yeah. I like that Ford deployed engineer because it's that top popular deployment of skills. You bring up a couple of different things. One is when you have diversity at the edge, a lot of connections, a lot of things are disaggregated, but still connected on the network. It's hard to write static rules to route and control. That's a great use case for AI intelligence because the AI could help move things around. So if I got a node here, a device.

Omri Yoffe

>> And the key here is to leave enough buffer for human judgment and human policies because you don't want the AI to become the regulator of the rule base that you're mentioning. So yeah, you want to build, let's give a practical example. If you want to drive a specific intervention for a patient, they finished an operation, they have a God forbidden oncology problem, and after they finish their appointment, you have multiple avenues they can take. This could be a rule based, but there are other softer human decisions that are in the loop here, if that makes sense.

John Furrier

>> One of the things-

Omri Yoffe

>> You want to put those guardrails for human judgment to kick in.

John Furrier

>> Omri, one of the things we're seeing obviously here, New York Stock Exchange behind me is the option floor, equities on the other side. I see the board all the time.

Omri Yoffe

>> Yep.

John Furrier

>> Tesla, Nvidia, Palantir, all those big names are up there. Databricks is not yet a public company, but they're in town. They're going to come in later today. The success formula seems to be ones who can see that, build that data platform, have data lakes like Databricks, but Palantir has an operating system thinking around how they do their business and been very successful. And you're starting to see more of that span out into the cloud's been there. So Amazon nailed that, AWS. So now you're starting to see the enterprise starting to think differently. It's starting to think like, "Okay, I need to have intelligence and I won't say surveillance, awareness, observability-

Omri Yoffe

>> Absolutely....

John Furrier

>> Governance, guidance, security, all there, but make it real time." I can't overemphasize that seems to break a lot of things when things just don't go fast enough. All those companies are building this intelligent layer and understanding like a system. It's not like a database and throw some nodes on it. It's like a connected system. What's your reaction to that? Do you see that as a step forward?

Omri Yoffe

>> Yeah, it's a necessity. I think people imagine that AI will become this self-autonomous-

John Furrier

>> Evil.

Omri Yoffe

>> Evil or not.

John Furrier

>> Oh, good and bad.

Omri Yoffe

>> Just with a fire and forget system, and it's never going to be the case with enterprises and definitely not going to be the case for healthcare and life science, and you don't want it to be the case. What you want to do is identify, underwrite the high impact use cases that AI can do the heavy lifting, but leave place for humans to observe. And in the agents world, I can give some practical examples. Even if it's not a human, it's an agent, you need a system, you need an architecture that will have web APIs and we'll know what to prompt or what to ask from a different agent as a system and not as a specific application.

John Furrier

>> So I have to ask you, one of the hot topics has been in the past week has been the OpenAI and Anthropic DOD piece.

Omri Yoffe

>> Sure.

John Furrier

>> I saw a memo from Anthropic CEO Dario Amodei and he also put out his own opt-ed around criticizing OpenAI and the DOD was trying to force him to give me your stuff.

Omri Yoffe

>> Correct.

John Furrier

>> And I had that scoop actually and broke it before he actually made that story talking to my friends in DC in covering a lot of the DOD stuff is there wasn't that he was against AI.

Omri Yoffe

>> Sure.

John Furrier

>> It was just that they wanted humans in the loop.

Omri Yoffe

>> Sure.

John Furrier

>> That was his key thing. There was also a thing about surveillance, but I think that was a red herring. The real story was autonomy, if you rush to autonomy and autonomous, you're going to get there. Now I also have sources inside Anthropic that told me that he knows a lot more of what's happening and what could go wrong. So I built this as context to ask you this because you were kind of pointing out some of the governance things.

Omri Yoffe

>> Sure.

John Furrier

>> You can do it right. He got a lot of props by the way. He got kudos because he stood his ground and did the ethical thing. So there's ethics involved. I know war time is a whole different ballgame, but healthcare has got consequences too. People die or deliver or die based upon the right care. So I have to ask you, you're starting to see AI and you're already seeing the signs that this is coming. What is the right way to think about it? Obviously he's being conservative saying guardrails are critical. Human in the loop was his pet peeve. I want humans to be on this. What does that mean? Obviously DOD a little bit different because weapons are involved, but healthcare people feel the same way.

Omri Yoffe

>> I agree.

John Furrier

>> What's your reaction and what are people like-

Omri Yoffe

>> The life or death implications in many ways. And I'll say two things and I'll be very blunt with your permission. Yes, please. First, this is irrelated to technology. I do think we live in a civil society. There's a reason we are the New York Stock Exchange. There's incredible people and thought leaders built those foundations for a reason. And I do think we as builders and thought leaders need to work with them to be able to blend ourselves within the right cultural ecosystem we're in. I don't see us as this AI tribe building autonomous systems. And I think it's critical specifically in healthcare to work with the government, with the largest health systems, providers, biotech companies, large pharma. And I'm very proud to say we're doing it with some of the most critical decision makers to make sure they're forward-thinking with their policies, number one. Number two, we are big believers that if you de identify, if you're being very protective of the data before it gets into the system and don't infringe privacy at the edge device, you're in a much better place. And I think that's the answer to most of your derivative questions to what you asked. So as long as you treat this religiously and saying, "Those are people's privacy and those are the most fragile health information, let's tokenize it and de identify it at the get go and only then bring it into the system. And then you can treat an anonymized patient and the AI can predict, AI can trigger actions, and the hospital or your provider will keep the information as easy."

John Furrier

>> Yeah, that's a really good point. I'm really glad you brought that. It's very nuanced, but I want to double click and just highlight that because just last week in Barcelona, I was talking with Mark Austin with AT&T, he's one of their scientists. He built all their top models and turns out OpenAI doesn't speak telcos. We actually donated to open source 19 models because there's a lot of configuration and jargon in there. So he donated all that. He talked about governance and getting that security right and the privacy because he says, "Once you get that right, the agentic piece kicks into high gear."

Omri Yoffe

>> Absolutely.

John Furrier

>> And his point was there's no silver bullet. You got to grind it out.

Omri Yoffe

>> Absolutely.

John Furrier

>> That's my words, not his. But he said, "You just got to do the work, but the benefits are significant." Your reaction.

Omri Yoffe

>> I grew up in a house that had some agriculture background and my parents, they have an incredible, a few acres of olive oil. And you have one or two times of the year, it's specifically one time with some other, depends how you count it, depends on the systems and weather that you basically get the olives. But before that, there's lots of heavy lifting and lots of tiny, anywhere from water to insects and so forth. And the analogy here, you must do the 95% of the heavy work you're doing. And if you label correctly, if you tokenize correctly, if you build the right integrations, API integrations to your CRM and to your supply chain and to your hospitalization beds system, then an agent could become incredible. I'll give a practical example. We work with the largest healthcare provider in the world. And basically we built technology with them and it took lots of time that we can tokenize every single patient they have. And it took a lot of time. But today we can take those tokens and because we have the ability to see when someone finished a specific appointment or surgery, you can now see all of the other areas that the same token did across their life journey, their patient life journey. So I know it sounds nuanced, but it's exactly what you're saying. You're doing the heavy lifting at the beginning and then an agent could be very effective with no compliance limitations.

John Furrier

>> It's a great lesson. I wanted to bring that up because the enterprises are all trying to rush into it like the DOD and others. So I wanted to highlight that. Let's talk about the edge. Okay, so you got this, you laid out the distributed computing platform, all the database work, the stuff you got to do and all the database. But now the edge is getting data, new data.

Omri Yoffe

>> True.

John Furrier

>> It's not like old data. It's like either instrumentation, some sort of context is coming into the edge device. There better be some compute there or the model better be there. Talk about the new data impact to the realm of other data that is going to be being managed real time. So you've got new data, how does that get trained, gets reasoned, do the next step action. You got to figure out what it is. Then you got to go do the intelligence, do the reasoning and then reinforce.

Omri Yoffe

>> So let's split two levels of, let's call it generated data. The first is, it's not old world, but this is, there's a handicap or a gated amount of info that civilization was able to generate over the last 40, 50 years hosted on the web. And therefore this has a sitting, right? And I'm sure you know it, LLM models are good as much as they can scrape the web and basically see what human wisdom and best practices and knowledge has created. This is limited. You're talking about, I think the new layer of intelligence and to your point, if an agent realized that the specific action needs to be made for a patient, you just need to make sure the reliability or the quality of that decision so you don't treat it as a fact, if that makes sense.

John Furrier

>> Yeah.

Omri Yoffe

>> Once you've built the guardrails to make sure it's a fact, it's a building block for other things that you can store and keep this insight with you. But there's a junction there that you want to make sure you don't create compound, non-truth seeking, let's call it system. But if you do have those generated true prompts, those are stored on the device level. All

John Furrier

>> Right. So let's talk about your business now. Talk about some of the momentum you have, exciting things you got going on. What's your focus?

Omri Yoffe

>> Sure. So we at Vi, we try and focus on ... We don't like to oversell our numbers, so I'll try and be as home point as possible.

John Furrier

>> Sure, please.

Omri Yoffe

>> First of all, we are here to build a profitable, self-sustained, thriving business for years to come. What I can tell you that we are well within the nine figure ARR position. The company is either on a breakeven or a profitable state over the last four to five years, backed by some incredible investors, General Atlantic, and some other excellent growth funds and strategics. And from a growth perspective, we are going to pass the 200 million anonymized or de identified million patients. It's the largest de identified data web in the world. We productize over 20 LLM and machine learning models across our full platform. We have tens of agents that are working at very large scale. When I say agent, it's not a specific task. It's an agentic system that can cover full workflow. And my goal is to continue building the business for years to come and to get to a billion dollar of total contracted value by the end of next year. So basically to have a predictability over the next three to four years, not necessarily an AR metric, but the ability to underwrite and be confident we can drive material revenue in a profitable way.

John Furrier

>> Well, you got some great momentum. I super appreciate what you're doing. I think healthcare is one of the most obvious but most important AI areas to be transformed given what's already in place. It's not like the existing incumbent systems are a liability. It's actually an asset. And so you could get in this nice system there, you get the data, you get AI infusion in there. So I have to ask you, how are you impacting, say, collaborating partners like there's huge breakthroughs. I'm going to be at Nvidia next week at GTC, and there's a huge life sciences tsunami of breakthroughs. And it's almost like the Cambrian explosion of breakthroughs because supercomputing has never been democratized. It's good.

Omri Yoffe

>> And by the way, this is for our next discussion, we are heavily investing into what you call, let's call it advanced computing, but given our vast amounts of data, the ability to leverage it into a stronger than a classic cluster compute power, very interesting combinational programs opportunity. And again, this is for our next chat, but to your question, I tell my team every morning and to myself, life cannot be about getting another paycheck or setting a company and going to a nice hotel or a restaurant and die. We're here to your questions on impact with partners to measure the way we make impact and in two to three to five years, look back and be very clear, did we use our most productive years in a meaningful way? So I'll give-

John Furrier

>> And you have data too. You have the ability to affect change.

Omri Yoffe

>> Correct....

John Furrier

>> faster that would have taken months of get a meeting, what data do you have? Let's do a pilot.

Omri Yoffe

>> Correct.

John Furrier

>> Now it's like, just boom, you're in.

Omri Yoffe

>> And I'll give two examples. Two practical examples. We helped over 50 of the most important drugs out there to accelerate. So we don't help on the molecule level, but we helped them be more precise and predictive, resulting in faster drug development. So you have specific drugs that are life-changing for families and people. And number two, we generated, saved or generated north of $2 billion to our enterprise partners, resulting in a reduced cost of care. So we need to make sure that we keep our eyes on the ball impact wise as well.

John Furrier

>> I got to ask you a question. I know we're running out of time. We definitely do another chat for sure on theCUBE, it's a great topic and a lot to talk about. My oldest daughter built out the clinical trials team at UCSF.

Omri Yoffe

>> Congrats.

John Furrier

>> Moved to New York. And she's looking at opportunities. And one of the things that she did was large cardiovascular trials, long process, tons of paperwork. Okay, things go to the market, they have donors, doctors involved for a teaching hospital. But the trend I'm seeing in companies that are in healthcare, whether it's an Oura Ring or these IoT devices, it's not just for the critical instrumentation for heart and other diseases, you start to see companies do set up clinical trial practices.

Omri Yoffe

>> For sure.

John Furrier

>> Can you share in scope, is that real? Is the idea of having accelerated trials? Because in IoT, my wearable that getting my heart rate could be much more better. It's not my heart, but you still got to get stuff approval. Are you seeing a different market developing there?

Omri Yoffe

>> We definitely see. So again, we need to stay very disciplined with our mission. We need to help those examples being precise, efficient, and ROI driven, but we do see incredible example for two things. The first, you see more and more solution teams, meaning, they leverage technology to become a service, an AI native service provider, if that makes sense so your daughter can decide without insulting any of the large CROs out there, which are great people. Some of them are great partners,

John Furrier

>> But which CRO?

Omri Yoffe

>> So it's basically those think about them as facilitators for large clinical trials.

John Furrier

>> Okay, got it.

Omri Yoffe

>> And your daughter can decide that she's going to be an AI native CRO and she can decide that she will wake up in the morning, she will still have some people working with her, but the default for most of the actions for finding the patients, for getting their clinical info, for analyzing the-

John Furrier

>> So she can be a service provider, knowledge.

Omri Yoffe

>> But it's a high margin one. She can become an incredible ... It sounds ... We are the New York Stock Exchange, so you know the difference between tech multipliers and service. She can be a tech multiplier, accelerating clinical trials by herself with instead of 1,000 people, maybe 70 people, but with many agents working for her. And this is something we see more and more in our space. So if I would recommend her, learn the domain, don't skip that. AI cannot ... If you're not a domain expert, you don't have an edge, but if you are becoming a domain expert, in her case for clinical trial, but you have AI native skillsets.

John Furrier

>> And so the output there, she would in this example, or anyone doing this, they're helping people get trials completed faster. That's the market?

Omri Yoffe

>> Yep. Very important mission, by the way.

John Furrier

>> All right, Omri, great to have you on. Super exciting. Again, we're kind of coming into the ... We're seeing on the tech side, data lakes, horizontally scalable, data with intelligence, now with AI on the edge, AI factories. Hospitals will be their own AI factory. All these data centers in Texas big for AI, great, great stuff.

Omri Yoffe

>> They're getting there.

John Furrier

>> Hospitals will need their own node. That's coming.

Omri Yoffe

>> It's coming. And again, I don't want to ... We have great relationships and partners on that field. As a patient, you're going to see more and more predictive, precise patient journeys, and we want to be a big part of it.

John Furrier

>> All right, great. Thanks for coming in. We appreciate it.

Omri Yoffe

>> Good stuff.

John Furrier

>> All right, great stuff. Healthcare is being transformed and AI can be a real force multiplier for good. A lot of new things. The market's evolving superfast. As new players come in, these platforms will enable more and more service providers, more and more better healthcare and just a better system. And again, we're healthier, we live longer, better society. We're doing our part here in theCUBE to bring you all the action. I'm John Furrier, host of theCUBE. Thanks for watching.