How This Cardiologist’s New Startup Leverages Powerful AI

Transcription PNC Podcast Episode 425

How This Cardiologist's New Startup Leverages Powerful AI

- Interview with Dr. Saurubh Gupta

John: Today's guest is a full-time practicing cardiologist, chief of cardiology and professor of medicine. But I invited him to the podcast, not to talk about those things so much, but to talk about his role as a founder and CEO of an AI startup. So with that, we'll get right to it. Dr. Saurubh Gupta, welcome to the podcast.

Dr. Saurubh Gupta: Thank you very much, John. I really appreciate you inviting me and I'm looking forward to our conversation today.

John: It'll be fun. I think we'll all learn a lot, including me, because I'm fascinated by AI and I don't know, I think I'm going to have to get other guests who have somehow leveraged their medical side to implement something in AI. So you're a good one to start with, I think. But let's start though, by letting our listeners know like about your clinical background, just kind of before we get into this new stuff that you're up to lately.

Dr. Saurubh Gupta: Sure, John. By background, I'm a cardiologist and I was a long career in cardiology in both academic and non-academic centers and various roles within organizations, director of the fellowship program. But my clinical focus generally has been in newer and cutting edge innovation, even within cardiology.

For example, my clinical focus over the last decade, well actually even more than that, about 15 years or so, has been in valvular heart disease and these cutting edge therapies where we can now replace heart valves through the groin. A lot in our audience have heard about transcatheter aortic valve replacement. So my team and me, we did the first commercial case on the West Coast way back in 2012.

And lo and behold, in 2019, that was a dominant way of doing aortic valve surgery. And within that experience set is also interfacing with various professional and public service organizations. So for example, with the American College of Cardiology, I just finished a term on the board of governors for that, as well as some work with the Oregon Medical Board in just understanding both the medical regulation side, but also be able to contribute to the regulation of our own profession.

John: You're plugging along, you're doing all these things, you're innovating in cardiology, and then you decided to try something different. Now, let's go back because I think we're going to talk about what you're doing now, but there was something that kind of preceded that. So why don't you tell us how you decided to maybe reach out to something more involving making a business or creating some kind of structure?

Dr. Saurubh Gupta: I think if I go and think about the arc of my career, I've always been really interested in innovation. And almost by definition, modern medicine is like flying planes. We want procedures and medicine to be a bit more algorithmic. We want people to be able to provide standardized care. And the analogy I use is that we don't want a pilot innovating when they're flying the plane. We just want them to fly exactly the same way they've flown for the last 5,000 flying hours.

And to some extent, modern medicine is at that stage. And obviously, innovation is important, all the advances in medicine are important. Unfortunately, the specialty that I picked, cardiology, the disease states are very prevalent.

And then in general, it has been at the forefront of innovation just because of the volume and the burden of disease. And so newer therapies, newer procedures have come out first and that's shaped my thinking on what really interested me. So even within the professional spheres, it was always on the innovation, whether that be on the healthcare delivery side, whether that be in these procedures that fundamentally changed the paradigm of how heart surgery works.

And then fast forward to, okay, where does this go next from a career perspective? And I remember having a feeling of having my impact be far more broader than we have in clinical medicine. I mean, if you think about it, there's absolutely the most wonderful thing is to deliver care to one patient at a time and then seeing the outcomes of that. But then how do you influence that at a scale that fundamentally transforms how we do things and how we look at things? So those are the seeds of the journey themselves.

John: Yeah, I think, and I've had this feeling myself, although maybe I've not pursued it as much, maybe at a small level, my wife had her own business, is sort of, you want to do something that really challenges you, at the same time might help a lot of people, whether it's local or even national or international. And if you're healthy and you're vigorous and it's like, okay, I've done that procedure a hundred times, 500 times, maybe I want to try something else. So when did you first start dabbling in something, let's say it wasn't quite the same as person by person care?

Dr. Saurubh Gupta: About 10 or 12 years ago, 12 years ago or so, I approached this as you and I might approach this from a professional level. I say, okay, this is something that I'm interested in if I have ideas. And then when I talked to some of my friends who have done this successfully at that time, I said, hey, I have an idea.

And they put a hand on my shoulder and said, Saurubh, let me tell you something, ideas are dime a dozen, it's the execution that matters. So that was a sobering experience. And what I took from that is that just because we happen to be at the top of our game in cardiology or medicine, we're all highly trained physicians and healthcare providers, doesn't mean that that same skillset translates to all walks of life.

And so my initial experience was with learning. And I approached this as if I was doing a second fellowship. Say, okay, I have to learn, I have to understand this ecosystem, I have to see what makes an idea go from a drawing board to an actual execution to then a product to then something that actually is a business.

And so I started off, the good part was that being in academic medicine, we're always approached by people, really smart people with ideas, whether that they were in the bioengineering departments or whether they were outside the university. And I took that opportunity to learn and fill the skill gaps that I know I didn't have. And then fast forward that to a chance, which is the COVID pandemic.

During the pandemic, at the very earlier parts of the pandemic, the governor of Oregon, the state that I am in, said we are short ventilators and people are dying. And if you remember back to those times, we saw people, for example, in New York, patients hooked up, several patients hooked up to the same ventilator, which was just astounding to see and like an experience that I'm sure none of us have ever had before. And so we said, okay, let's make a ventilator and let's make a ventilator company.

And that was the genesis of that company. We ran it like a startup with very fast execution, fail early, fail fast, get up and move on. And our core principles were, one, we want to do this for the goodness of humanity, two, we don't actually want to take away from the existing supply chains.

For ventilator manufacturers, as you might recall, supply chains were a huge challenge at that time. And then third, that it needed to be clinical grade and open source, meaning anybody could use it, anybody could profit from it in a non-business way. And from that point on, we actually got it to a level where it was FDA authorized and got deployed.

That was a transformative experience in the leverage and the power of teams and this whole new concept of asynchronous working. But before that, the fundamental models were, okay, everybody goes to the office and we work.

John: Yeah, I don't know, you just get bored maybe a little bit and then you have all these other ambitions and it's like, okay, well, there's plenty of problems to solve. So tell us about the most recent one, the company you're now running, because that's a whole discussion in and of itself, but it'd be interesting to hear how you came up with the idea for that and eventually put that in motion.

Dr. Saurubh Gupta: Totally. As we were transitioning off of LifeMech, which was the ventilator company, there's a lot of people who came up to us or came up to me and said, well, we should make business out of this, right? And there was some very terrific technology that we developed.

It was absolutely, we changed the way how ventilators were thought of. And except that none of us felt right about commercializing that technology because these were engineers from Intel, Apple, Microsoft, Boeing, who were working as part of that team and giving thousands and thousands of their hours of their time free. So, and then about the same time, we started seeing the trends emerge within that group even.

My chief technical officer actually is a dear friend, but from that experience around this emerging artificial intelligence. And it's a very nebulous term. What does that even mean? Can intelligence be artificial? But immediately it was clear to us and to me that this has broad applications and transformative applications across broad swaths of what we do as a society and fundamentally within medicine. And then the problems were right there in front of us.

For example, medical data, how disorganized that is, how labor intensive it is, how hard it is to detect signals. And our fundamental premise was that what makes a good doctor or a good healthcare provider good? It's the judgment and the wisdom and the insights, those very intangible skillset that perhaps to some level can be taught, but we think that that's where judgment is.

And how do you enable these modern systems to reach that level of sophistication and where are the interfaces with that? And then from that evolution on came to, okay, what problems are we trying to solve? And I think the key differences between being a physician founder versus let's say a technical founder is that sometimes, not all the time, technical founders will come up with some really cool ideas, some really cool technology, and then they approach a problem set to solve.

And what we wanted to do was, okay, here are the problems, here are the problems in medicine. And as you know, some of even the simpler problems are not solved yet. And how do you then develop technology that solves those problems?

And that's how we got started into CoreMetrics on taking medical data and trying to make sense of it at a very broad level. And then from there, it evolved into a platform which has a very, very robust architecture in the back end that supports use case scenarios. And I'm sure that we'll talk a little bit about Verix, which is a medical legal software that is now out in the marketplace and has adoption among attorneys and medical expert witnesses.

John: Well, I would imagine that, the question you were trying to answer has been around for a long, long time. And I can imagine if it were me, I would be looking and saying, but now with AI, it sounds like we might have a way to actually address it because to just write reams and reams of code to try and have every algorithm and put every, because they've been doing that for years and they haven't really gotten that far in terms of judgment and decision-making and diagnosing some obscure things. Is it true? Is that what you saw? Okay, AI is something different. It's probably something that would apply to this. And then you had to pretty much jump into that, I would imagine.

Dr. Saurubh Gupta: Now that's exactly right. And you articulated just beautifully on what the problems are with AI. And even though in the last three years or so, since let's say, ChatGPT as its first large language model came out, we've made astounding progress. But some of those problems are not solved problems yet.

Problems such as patent recognition, artificial intelligence has been evolving. It's not aha moment, right? This is built on a series of advances over the last decade or two decades, advances in machine learning and advances in neural network analyses. So AI does really good where there are patents.

For example, radiology, pathology, those are fantastic in early test cases because it really does do a good job and it takes away the bias. But when it goes to the judgment and the wisdom, that's a much, much difficult problem to solve. How do you even quantify that? What constitutes good judgment? What constitutes success? And I still think AI struggles at that. And what I also am a believer of is that, that healthcare professionals are ideally placed to be able to solve some of the bigger challenges in healthcare.

Obviously it has to be a partnership between technology and the really smart engineers who are coming out with these therapies, but there is nothing more powerful than actually having lived the experiences and lived the problems that you're trying to seek. And then in the startup world, they call it founder market fit. And obviously the next term that comes from that is a product market fit or PMF. Those are some of the themes that I apply around on what problems are we solving?

How are we solving it? And what tools are we using to solve it? In that sense, I see AI as a tool and you're absolutely correct when you mentioned a few minutes ago that these tools are evolving. And I wouldn't even say a year's time, they're evolving at a weekly cadence. There are strategies and tools and techniques that we are using in our own software, for example, that actually did not exist three months ago. We are that advanced and that is how advanced these technologies are, but how fast they're iterating. So we are basically at a fundamental level trying to solve some of the problems.

John: Well, it sounds like the experiences you had with the ventilators taught you a lot in terms of working on a team, leading teams, managing things. Yeah, there's a whole list of skills you need. So let's get into the nitty gritty of your current company. You started this two years ago. What has been going on? What part of it do you run yourself or oversee and how much do you have to employ others? Do you have many other staff? Are you using part-time experts and like programmers? Just kind of give us a picture of what you're actually doing.

Dr. Saurubh Gupta: Yeah, I think starting a company is a completely different experience than having an idea. It clearly requires a lot of partners. It requires a lot of people around you that help you succeed. And I was fortunate to have had this opportunity during the pandemic to work with a really, really talented group of people. For example, my chief technical officer is one of the smartest people I know. He's an engineer and I'll brag a little bit about him.

He's the guy who did the drone show at Superbowl a few years ago. He ran the entire team and he understands this at a completely different level. And from that point on ideation on whether this is a problem worth solving, and do we have the tools and the skillset, excuse me, to solve it.

Fast forward to that, then the initial efforts around this are generally self-funded. So it takes some self-belief, some commitment to say, okay, I think that we are on to something. And then the next step forward from that is raising capital.

In the beginning, the VCs actually are not the way to go as you probably know. Some of them will take a bet on very early platforms, but most of the time it's family and friends and personal investors who get you off the ground. And now we are about a 16 member team.

We have different roles. I have a chief operations officer who does the day-to-day operations pieces on what's working, what's not working, how do we make this optimized? We have a chief compliance officer.

In fact, just last night, we heard back that we are now SOC 2 compliant. And for us in the medical world, that may not mean much, but if you go to any tech website, you will see the words SOC 2 compliant. And what it is, it's an external auditor that looks at all your systems and controls and operations and your software and talks about vulnerabilities, looks at how your defenses are structured, what does your penetration testing look like, essentially the safety and security around it. And we have a chief product officer, and then we have several engineers that we employ.

John: Now, of course, we spoke before the podcast started and I know that your company CoreMetrics, which I think anyone can find if they look online right at coremetrics.com. One of the things it's focusing on is the data review and analysis and reporting as it relates to, let's see, it's medical legal. So explain exactly what that is and why you thought that was a good first start or one of the first places to start.

Dr. Saurubh Gupta: Of course. So as I mentioned, we have conceptualized CoreMetrics as a platform that can support multiple verticals across the healthcare domain. And one of the problems that we've seen is how messy medical data is. And obviously it is very evident in clinical care, but then it's also very evident in the medical legal industry. And I know that people talk about lawyers in one way or the other, but I think the law is an important concept in all of our life in modern democracies. And sometimes patients do get harmed.

We felt that that was a good application of our software to start solving that problem. So essentially we can have attorneys and expert witnesses and legal nurse consultants who get these stacks, sometimes hundreds and thousands of pages of data, and it can come in different formats. It can come in different types of notes.

We've created an integrated platform that just takes all of that data, including imaging and any sort of raw data. And then we basically can provide judgment and insights to both the attorneys as well as the medical experts. So for example, some of the features you could do is you could talk to the software.

You can literally talk to it. It has an audio button and say, tell me about this patient. And we'll come up with a fairly incisive answer. What happened? Just those types of questions to fact-finding questions like, can you show me a trend of the blood pressure in the last one hour before their surgery or the first 36 hours after the surgery? And historically that would take a lot of digging into the datasets to get that type of information. Not only is that cumbersome, but it's prone to error and fatigue. And as a company, we don't even believe in AI. And I know you're smiling there to say, why do I say that?

Because I do believe that these modern systems are best utilized as collaborative intelligence rather than artificial intelligence. So our hope is to collaborate and develop AI systems that collaborate with humans, with human-in-the-loop design, as well as amplifying the power of the human mind and the judgment and business. Will they evolve to having some insights and some wisdom of their selves? Absolutely will. But I still think that the fundamental outlook, our fundamental outlook on this problem is to be collaborative with the intelligence that we're developing.

John: You have clients using this currently, and some of my listeners are expert witnesses, some do other kinds of medical legal consulting. I don't probably have any outright attorneys listening very often or large firms. Who is this made for? Because I have no objection to promoting it if it's something that works and it solves a problem. So tell me who might be interested in this.

Dr. Saurubh Gupta: Totally. Clearly law firms, especially on the personal injury and medical malpractice side, those are some of our customers. As well as from the expert witness sides, whether that be physicians or nurses or legal nurse consultants, those are all customers. And in fact, we have several of them that are using our software as customers.

John: Interesting, because not only I have some expert witness listeners, but I also have people who teach expert witness or teach medical legal consulting. And I think they would be very interested in hearing about this because it solves one of their problems. It's a massive amount of data they got to go through.

And to be really fair and comprehensive, they have to sift through it all. And a physician might delegate it to a nurse, but still, the physician still has to go through a lot of it. And if they could solve this problem of even just being able to query and ask the questions, like you said, to just answer questions as it pops in their head, not have to go back and read everything, that would be huge.

Dr. Saurubh Gupta: Absolutely. And I think the toughest part of all of these was actually to build in the compliance framework, believe it or not. And so not only do we have to solve the problems as you just articulated, but you have to do it in a compliant way because essentially health records and legal records should and do in our system operating is what we call a zero trust environment.

Basically, trust nobody, believe nothing. We feel that it's a very powerful tool. Now, the interesting part is that when some of the expert witnesses and people who are new to this whole field of applications of AI into medical datasets, they will say, well, I'm searching in the data, but it's not quite the search the way that we're used to thinking of search.

It is not looking for, for example, in a traditional medical record, let's say you search for the words ECG, and it would give you 12,000 references on where the word ECG appears. It could appear in the note that the ECG was done. It could appear in the orders that the ECG is ordered.

Whereas you really say, show me the ECG, which is what you really care about. Or was an ECG done within five minutes after the patient presented with chest pain? Our platform does that in a very intelligent way.

It doesn't really look at it from a keyword search. It looks at it in a much more holistic semantic way, going into context and going into intent. And that's the part where we continue to make very rapid progress in giving more context and more intent on what exactly is the intent of that question.

John: Not only are you doing something new for yourself, but the thing you're doing is also something really new, you know, a cutting edge, like you said, in the cardiology realm, you were often doing some cutting edge procedures. So there's other people I've spoken with, and they're really interested in doing something with AI. And so I guess I'm going to shift to maybe asking that question.

If you have someone like yourself in whatever field of medicine, and they want to, I guess they still need to figure out the problem they're going to solve, but what advice can you give them to help them understand, you know, whether AI is the solution or how to understand AI if they're thinking about going that direction?

Dr. Saurubh Gupta: There are a few themes that I have learned over the years. You know, I say that my superpower is being a lifelong learner and asking the right questions. And advice number one would be to ask the correct questions. Is that a problem to solve? Because there are a number of cool technologies, technologies evolve, and we see technologies as toolkits and tool sets that continue to evolve. Problems remain. Is it a problem worth solving?

Second, what I found early founders, people who come to me for advice, to be very secretive in the sense that, I think a natural impulse for all of us is somebody is going to steal my idea. And I hate to break it, but most of these things have likely been thought of. Talking to other people around me, people I trusted, of course, made me a much better founder, being the ability to refine the problem statements, refine my own thinking on what we should build, how we should build that. So mentorship is key. I would advise identifying people that you trust in your trust circle.

And generally, a lot of founders are actually very open. I certainly have benefited a lot from people in the last 10, 12 years who very voluntarily and graciously gave their time. They still continue to do that. Some of my mentors right now are people who just invest time and energy into making me successful. So identifying that is key. I think we're never too advanced or too old or too far ahead in our fields to ever lose the need for a mentor, especially in areas where we have skill gaps.

Third, especially from healthcare perspective, is to be honest about the skill gaps. And we have several. I certainly still have a lot of skill gaps and see how am I going to fill those. Before the podcast, you were asking me about the technical aspect of it on how in depth do I understand the algorithms and the training and the models. I think my understanding is evolving. I can now write simple code.

My team tells me it's garbage code, but I'm still proud that I can write it. But I do believe that there are skillsets that I've picked along the way that I understand this a lot better than, for example, I did even a year ago.

John: Well, let me ask you this question, kind of a leading question. But so as you were thinking about this problem and the solution to this problem, and of course you had all these people in mind you had already worked with, but you're learning by learning from other people. I mean, sure, you're reading and doing that, but you didn't say, well, I need to go back and do a year's course on whatever, some related topic. You just said, I can learn this stuff fast enough, easier talking and being mentored and coached. Is that what I'm taking from what you said?

Dr. Saurubh Gupta: Yes and no. I have had mixed feelings about that, John. In a way, I think there's definitely acquisition of skillset, but then at a deep fundamental level, we are building something new, something that doesn't exist. And it certainly doesn't exist in the way we're building it. So one part of me always thinks about that what course can teach me to build something that hasn't been built before? And so certainly there's absolutely a fantastic educational program that help physicians and other healthcare providers understand AI.

But my view on this is that I will never be as proficient at the cutting edge technical tools that are being developed on a daily basis as some young kid out of Stanford. But I do have context and domain expertise and the ability to understand how to utilize that tool and a rudimentary understanding of what the tool is doing and how is it doing that. So certainly I think it's not at all unreasonable to take some courses that bolster that level of understanding. But beyond that, I think you just have to jump in.

John: Yeah, and similarly, you didn't go back and get an MBA, I don't think from what I know about you. You don't need an MBA to start a company, obviously. Some people feel like getting all those letters behind your name make a difference. But fine, if you want to do that, it's the course of your life and your career, but anybody can start a business, but you need to be open to learning and curious is what I'm getting from what you're telling me.

Dr. Saurubh Gupta: Totally. I think curiosity. And I'll share a thought experiment that we had when we were starting these new therapies on valve disease, right? Traditionally, valve surgery has been done by surgeons who cut open the chest. And then when this field came around, the obvious question was, who should do this? Should it be surgeons? Should it be cardiologists? And the same concept exists in multiple other therapies where as more and more things are getting less invasive. And my thinking on that is that that's not the correct question.

The correct question is who has the skill set to do it, not what specialty or what letters behind the habit. So think of this as you're a cardiologist, you should be able to read an EKG. In my mind, the correct question is, can you read an EKG? And then if the answer is yes, you could be an ER physician, you could be a physician associate, you could be a nurse practitioner, you could be an infectious disease doctor. The idea is to move to that type of thinking. So that's how I've approached it.

John: Yeah, I think it's worked very well for you so far, and it's good advice. Well, we're going to run out of time here, I think. So again, tell us how we find your company. And I guess I remind people that Saurubh is on LinkedIn. So if you want to just do an informal note or something or connect with them, you can do that. But again, the website for the company is what?

Dr. Saurubh Gupta: It's coremetrics.com. And obviously, very accessible. Feel free to reach out. I mean, I don't know a lot about a lot of things, but whatever I do, I'm very happy to share. And then, you know, listeners of your podcast, if they want to try our software, we're totally happy to give them a yearly subscription for free.

John: Oh, that'd be great. You want to give us any sort of hints as to some of the new things that coremetrics might be getting into? Or you want to just leave that off the table for now?

Dr. Saurubh Gupta: Oh, there's so much exciting stuff. So we have a major strategy meeting tomorrow night. That's Sunday night. I think not tomorrow night, so two days from now, where we're thinking of a very broad application of how and where we go next. But I think we'll have to leave that one for perhaps the next time you invite me back, if there's a repeat invitation. But it's something that's very, very exciting and far, far broader impact.

John: Good. Well, that's exciting. Yeah, that's a good teaser. I'll have you come back once you have that locked down. All right, Saurubh, that's great. I think we'll leave it there. I appreciate the time and the wisdom you've shared with us. It's exciting. And we'll hear more in the future, hopefully, with that, and I'll say goodbye.

Dr. Saurubh Gupta: Thank you, John. I appreciate you having me on your show, and it was just such a pleasure having this conversation with you.

John: Bye now.