99 - From Wall Street to Healthcare | Guest Sergei Polevikov on AI, VC, & Transforming Patient Care

Show Notes

In this episode, Vic hosts Sergei Polevikov, founder of Well.ai and author of the Health AI Uncut newsletter, to discuss the evolution of AI in healthcare and finance. Sergei shares his journey from Belarus to working at the Federal Reserve, exploring how his math expertise led him to Wall Street, where he focused on AI and machine learning. He explains the challenges he faced starting Well.ai, aiming to streamline patient-clinician communication. They delve into the barriers to healthcare innovation, the hesitancy of providers to adopt new tech, and the role of regulatory bodies. Sergei also critiques the venture capital model in healthcare, contrasting sustainable innovation with hype-driven investments, and emphasizes the importance of practical, patient-centered AI applications.

Sergei Polevikov Bio:

Sergei Polevikov is a Ph.D.-trained mathematician, data scientist, AI entrepreneur, economist, and researcher with over 30 academic manuscripts. He co-founded WellAI, a digital health startup, and is a member of the IFCC’s Working Group on Artificial Intelligence and Genomics Diagnostics (WG-AIGD). 

Sergei also authors the popular and provocative Substack newsletter “AI Health Uncut”, offering brutally honest insights from a healthcare AI founder, based on his experiences, mistakes, successes, and perspectives.

Connect with Sergei

• Company Well.ai: WellAI.Health
• Substack Newsletter: AI Health Uncut
• LinkedIn
• X

AI Health Uncut articles discussed:  

The AI Bubble? It's Really a Startup Bubble, and It's Popping Big Time

The Health AI IPO Checklist: How to Spot the Next Unicorn or Sniff Out the Next Donkey

HF Site

YOUTUBE
INSTAGRAM
TWITTER
FACEBOOK
LINKEDIN
TIKTOK

What is Health:Further?

Every week, healthcare VCs and Jumpstart Health Investors co-founders Vic Gatto and Marcus Whitney review and unpack the happenings in US Healthcare, finance, technology and policy. With a firm belief that our healthcare system is doomed without entrepreneurship, they work through the mud to find the jewels, highlight headwinds and tailwinds, and bring on the smartest guests to fill in the gaps.

If you enjoy this content, please take a moment to rate and review it.

Your feedback will greatly impact our ability to reach more people.

Thank you.

Okay.

Welcome to have further.

We have a guest show today.

Sergei Volkov.

Thanks for doing this.

Really appreciate it.

Sergei, you are the author of a really impactful newsletter called Health AI Uncut and also founded, you're wearing the t shirt, you have the t shirt to prove it, founder of Well.

ai.

Thanks for doing this.

Appreciate it.

No problem.

Thank you for inviting me, Vic.

So as we start off, maybe give the audience a little bit about your background.

You were born outside of the U.

S.

and then came here and just give us your sort of personal story.

Thank you so much.

Uh, yes, thank you by my accent.

You can probably figure out I'm from outside the U S I'm from a small country, Belarus, middle class family, which is by American standards, probably very, very poor family.

Uh, but yeah, I had a good background.

Both my parents, um, came from science from academia.

So I had very good science and math and physics knowledge, uh, graduated.

Bachelor's, um, in Minsk and, uh, kind of looking for applications of my math knowledge.

And, um, when I started looking around, I figured that a lot of topics in economics and finance, they were, um, math based and evidence based, and that kind of intrigued me.

And I started looking into grad schools, because at the time there was, at least, From my perspective, the most probable reason to get to a good school and maybe, uh, get a good job.

So that's how I ended up in the U S I, um, went to a couple of graduate programs here, uh, at SMU, could have been in Texas for a couple of years, actually, uh, while being at SMU solid Methodist.

I've met, um, uh, some interesting folks from the Federal Reserve Bank.

And in fact, the chief economist at the time, Mike Cox, uh, eventually became a boss.

So I spent, um, I spent a few months on that and kind of continued working remotely, uh, for Mike.

And in fact, it was, I think, pretty, instrumental contributor to his book, uh, myths of the rich and poor, which is kind of in line with my article that I posted today, which is basically, you know, why is America so great in innovation and, um, um, you know, what has gone wrong potentially in the last decade or two, because that book actually came out in 1999.

So that was.

Yeah.

It's like a

dot com, like run up to the internet craziness.

Yeah.

And, you know, a lot of the topics that were popular back then, you know, we've just started getting smartphones and, um, not smart smartphones, uh, I'm sorry, mobile phones and a lot of improvements in technology and obviously internet.

And that's how this whole conversation started.

But the book is actually a very good overview of just, you know, historic wisdom and knowledge on innovation.

And, um, yeah, so I, I was kind of like a stats guy.

So there are a lot of graphs.

How

was it to come from Belarus and then be working at the U.

S.

Fed?

That must have been an interesting transition.

I was surprised they hired me, you know, with, especially with a name like mine.

Uh, but no, it was, uh, it was a lot of fun, especially given that it was my first job and I was still kind of, being in grad school.

So I started as an intern and then kind of developed into something more and spend, um, actually.

In total, almost like two years, um, on and off.

So, so great.

A lot of great economists.

And, uh, I don't know if people realize, but the federal reserve banks, the research departments are basically like your econ departments and top, um, academic program firms like NYU, Stanford, Harvard, um, you know, some top economists, they.

Prefer sort of hands on experience, um, hands on experience on the markets and inflation and various surveys.

So they prefer actually being in the Fed as opposed to doing some research in academia.

So really, really top, uh, top of the crop, uh, I would imagine you

have access to the best data sources anywhere at the U S Fed.

Oh yeah, yeah, absolutely.

I mean, just having the databases and just looking at them all day long and, you know, exploring and, you know, getting some insight.

It was just like, for me, like being at that, like in my twenties, that was, that was paradise.

Yeah.

And then you went to the, the big times as far as like, uh, macroeconomics and tech went to Wall Street and really spent some time there, right?

Yeah.

Before that, I actually got my MBA from the University of Rochester, uh, where I actually, I was lucky to get a scholarship, which is not very common for, especially for grad programs, for MBA programs.

But, um, yeah.

And then after my MBA and I.

On top of that, I've always kind of been curious in the latest research, so I ended up, um, in every program I feel like I went to, I was actually friends with PhD students as opposed to, um, MBA students, which is kind of interesting, but I felt like this is where they cutting edge research was going on.

So yeah, I got my MBA, uh, got exposed to some cutting edge research in finance and ended up on Wall Street as quant analyst in the beginning, and then, um, developed into more portfolio management role, uh, but always been around math, around statistics, around machine learning.

And, uh, that was the time where the AI algorithms and AI training became popular.

And, uh, I was.

All kind of, um, deep into that and, you know, developed a lot of my own algorithms.

Um, I was, um, in particular, I did a lot of research on explainability.

Um, you know, when you are a fiduciary working for clients, obviously they're asking, you know, what is, what the heck is this black box model?

Can you give us more insights where all this great numbers coming from?

And, you know, it's kind of funny exercise because you're kind of, uh, trying to find a story.

So I feel like on Wall Street, it's a lot of a storytelling, but in all honesty, you don't always have the answer.

So we have all this great algorithms, but the, I think the biggest myth is that.

Um, you know, we can kind of, uh, explain the output, so to speak, and show the clients, you know, where everything's coming from, which is, uh, to be honest, in most cases, it helps

people feel good, but it's not necessarily accurate.

I mean, the data is the software algorithms are doing what they're doing.

You may have a predictive model, but that's different than being able to explain it.

Correct.

There is a, you know, people don't realize, I think, uh, even, you know, The ones that are dealing with AI models, but, um, you know, as I've learned over the years, there is no, you know, silver bullet there.

There is, there's always a trade off.

You, you have a great model that you can get some, you know, great, let's say accuracy or whatever the performance matrix is, but that's not always obvious why, where the results are coming in.

Or you can have a model that's highly explainable, but You know, it's not going to perform well.

So those are the trade offs you've been dealing with.

And over the last couple of days, I guess I would say there have been some, especially in academia, there have been some good research that are kind of trying to converge those two areas and sort of basically make some, some kind of, uh, Uh, golden middle, so to speak, uh, so, you know, that, that could be a good model, but also, um, you know, we not completely a black box, um, but again, it's, it's, it's, it's not an easy task.

It's, it's a, it's a very hard problem to solve.

Yeah, so the reason I was really interested in having you on, you have a great newsletter that we'll cover in a minute, but Wall Street was really an early adopter of machine learning and AI in 2004, five, and then through the great financial crisis.

And you were in the middle of that where, where the really kind of on the, on the frontier of how we could use AI in a business application.

And now, of course, after, uh, the transformer models have come to the mainstream, now every industry is looking into it, but you're one of the few people that have had decades working in this space.

So talk to me about, uh, in 2019, right before the pandemic, you founded Well.

AI.

And I don't, I don't know if that was.

It's a great time to start a company or a terrible time to start a company, but tell me about the founding story.

What were you trying to accomplish with Well.

ai and um, how to, you know, what, how's it evolved?

I mean, I, I'd be honest, it was, uh, almost, um, an accidental path for me.

Uh, and I'm, I'm very glad I ended up here.

I think, uh, all of us at some point should be thinking, uh, you know, not just about some financial outcomes, but also how we can get back to the community.

And I think that's how it all started.

Obviously it's a, it's a business, so we try to make money, but then when it's come combined with the idea that, uh, there's so much wrong with healthcare industry and, you know, being the best democracy in the world, and unfortunately I've experienced the lack of democracy.

So I know how great America could be.

I was shocked how bad things were and still are in healthcare.

Um, so I, it was actually my company.

You know, that I worked in finance was restructuring.

So I was, um, actually thinking of what my next move is going to be.

And, uh, one of my old friends, Daniel, who, uh, I met years before that, um, he's been involved in.

You know, various, uh, uh, bioinformatics projects and just, uh, you know, kind of did some work related to healthcare on academic side, but also on machine learning side.

So he kind of, uh, had this pitch to me that, um, he thought it would be great for the two of us to apply our AI skills to essentially help healthcare in some way.

And our way was basically.

Essentially, initially, I would say we were trying to build a better MetMD, uh, so to speak, um, after, after WebMD, uh, You know, we, we thought that with the greatest latest developments in AI, we could literally put the, sort of the real time information into patients hands.

So it was this kind of big vision.

One of the things that I heard, I don't know if it was in your newsletter or in one of the podcasts you've done previously, that you really, uh, draw an important distinction between, The caregivers, the doctors, nurses, other caregivers who are doing the best they can.

They're, they're great people.

I mean, they're doing incredible work, but we're not really empowering them.

We're not giving them the tools.

And so that causes suffering for the caregivers themselves.

And of course, for the patients because, um, so that distinction of not Um, not really blaming the people that are there doing their best work every day, but trying to bring them tools.

Is that a fair way to characterize it?

It's fair, and you have to understand the Uh, standpoint of clinicians and just medical community as a whole, they've been for four decades, kind of, uh, uh, the, the receivers, all this endless sort of it promotions, you know, starting with, uh, electronic health records and every little it vendor basically bombarding them with their little adults.

And that's sort of the problem that, and in fact, the book that I mentioned in, in my article this morning, uh, by Dr.

Jesse Funk, that was released today on Amazon.

Uh, he talks a lot about that.

It's basically, and you know, the, the reasons are, Very deep and complex, but the bottom line is that all the IT vendors in healthcare we have right now, or not all of them, but most of them, they're solving like little local problems.

Like none of them are addressing big, complex healthcare problems.

Little

point systems, maybe.

It's a little, it's a little hidden here and there.

A whole bunch of little things.

Yeah.

One doctor that I know very well once told me that, you know, I open my computer, I have.

apps.

So my first question to whoever I talked to, uh, you know, he's, he's a very kind of nice kind of man.

So he wouldn't say no to anybody on the phone, but he, so his first question, why do I need app number 101?

If I, if I only have a hundred apps, when I open my computer on my iPhone, right?

So, so the problem is that we.

We sell apps.

So we, instead of having this, uh, solution that kind of global and addressing the healthcare, every little startup, they have their own add ons, you know, for every little problem.

So we have your, your script, you know, this end in scribing that kind of addresses one issue, one very little tiny solution, which is nice.

You know, then you have somebody else, maybe.

Try to, um, you know, API to an HR system.

So every, um, little solution is fragmented.

It's, it's, uh, it's, it's not integrated into some big solution.

And this is exactly what Dr.

Funk is talking in his book.

And that's part of the reason why we ended up here.

And there are other reasons as well.

But, um, and so when we started, well, yeah, it was actually, obviously we had this great vision that You know what, maybe we can sort of address the big issue, sort of, um, Getting as much information as possible to both clinicians and patients.

At the time we felt like nobody really did it in an efficient AI based manner, sort of seamless, and we thought we had a chance to develop something.

Uh, but then we kind of faced all sorts of problems that I can talk if you want to.

Yeah, well, that's the original vision.

And then, of course, uh, as you said about building it.

I would position it as learning, like what did you learn and what insights, uh, did you uncover that, that I could learn from, or the audience could learn from?

I've learned a number of lessons.

Number one is that being a newcomer to the industry, It's, it's a very hard sell, which is kind of obvious lesson, but, uh, especially for us, you know, both my partner and I, my business partner, we came from different industries, you know, finance, fintech, and software healthcare.

So to establish that rapport, and you may have the best product in the world, but if you don't have sort of the, the networking aspect.

I feel like you, it's very hard to succeed in healthcare.

So you need time, you need networking and it, it takes a lot of time.

So that's the first lesson.

The second lesson is that, um, even though I mentioned there's so many healthcare AI and digital health apps, there is still this, um, this kind of Adverse reaction from especially big players, whether it's payers or hospitals, when you try to tell them how this particular product can help you.

The, the attitude that we faced, especially initially was that, you know, you guys like five guys in a basement, you really think you can help us Cigna, you know, build something in data science.

We have 200 of our own data scientists.

Um, and that's, that's the true statement.

The problem is that They have thousands of other projects.

This may not be on top of their list.

We already have a product and you know, all we're asking is that, look, you can test the product, it's already there.

Um, but they have this attitude that they know best.

And of course, after years, I've been kind of following the, the marketplace, um, Almost none of them kind of built what we were kind of talking about.

Uh, it's the, the, the products, especially on the AI side that this big corporations they're building is kind of like either they have some issues and they have to be shelved or they just, the bureaucracy is so high there that they just ignore this, uh, whole market altogether.

So would you say like a non invented here culture?

Or more of a, uh, you know, insider with all the acronyms and reimbursement things, but what was the biggest source of the problem?

I think it's the perceived market power of some of these corporations.

In other words, I feel like once you get in some kind of power position in the industry, you know, you, you prefer the status quo, you know, you, you, uh, you look at your cash flows and your revenues and you're like, well, why, you know, why, Break something that's, that's not broken, you know, um, and I feel like a lot of this corporation, they have this attitude, um, and that's the reason that in my opinion, we haven't seen much innovation, especially in AI and digital health in healthcare industry, because, um, in my view, I don't know if you want to call it monopoly or monopsony or I guess oligopoly, because it's not just one company, but in certain industries, My opinion is that, um, there is the Oligopoly power and they wouldn't let newcomers and innovators through, um, because they, this is how they do things.

They, they, they, they're used to doing things in particular manner.

And I think once you get that power, you were actually, it's not in your interest to be innovative.

Yeah.

So that's, that's where we are.

Got it.

Obviously, I run a VC fund in Nashville, and that's one of the reasons we started here in Nashville and didn't move to New York, Boston, Silicon Valley, because I think in healthcare, particularly in healthcare, the, uh, you know, human life is at stake, much more so than in other industries, and so that means that the regulatory bodies is.

have a lot of power and, and will flex that power.

That probably is good to protect public health, but it also can be overused or hidden behind to protect other things that aren't necessarily for the best for the patients or, or the providers or the payers really, anyone.

I agree.

It goes both ways.

Um, I think the history of antitrust has shown that in my opinion, there could be some positives when.

Company or corporation gets too, too big.

Uh, like we've seen in the history with, you know, with, uh, with, uh, um, what's the Bell, AT& T slash Bell.

And, uh, where's Microsoft, you know, they're broken up and actually new innovations came out of that.

Uh, but I feel like in healthcare in particular, I don't, I don't see the regular to reaction.

The, the FTC and department of justice.

At least from what I'm reading, they're going into some regional and mid sized players, you know, which I guess is justified, but I wonder why in healthcare they're not going after, after the big guys.

So it's kind of like, in my view, kind of.

very misplaced, the effort that they've been making.

Um, I'm not trying to say it in their job, but I'm just kind of questioning what exactly, you know, why they're going to, you know, some local, after some local players, as opposed to addressing the big issue in the industry.

Yeah.

I'm not sure about the FTC regulation, but as far as like HHS and CMS, FDA, it's, uh, it's sort of like incremental change.

They test things.

And then as soon as.

you can see that it's safe and effective, then they will begin to allow innovation in.

So, sitting here in Nashville, we're able to work with, uh, the big provider groups and really just try to understand what do they want to make improvements on in the next 6, 12, 18 months, uh, as opposed to just sort of changing everything whole cloth.

Not that I don't think that would be valuable, I just think it doesn't really work so well in healthcare.

Silicon Valley has several times.

Tried that and it, it gets stuck in the regulations because it's human life's at stake.

So tell me where it, Well.

AI is right now.

You're also still on the board, uh, but you're no longer, um, CEO running it anymore.

Is it still around and is there a product that we could test and talk to the audience about what, uh, checking it out?

Yes, absolutely.

Our initial idea was to give it.

It's directly into the hands of patients.

Uh, but over time we evolved and as you mentioned with regulations, it's really tricky to, um, you know, to do direct to consumers.

So now we work directly with business.

Mainly, uh, medical practices and hospitals.

Um, I think the best way to test a product is to sign up for a demo on our website, will.

ai.

health.

Um, we have, well, at the top there in big letters, we have the demo scheduler.

Yeah.

So I'll link to that in the show notes.

What is the most common use case or what, what, uh, what do people expect?

What, uh, what does it deliver for them?

Say I have a medical practice with, uh, 30 docs.

Is that the right profile?

It's a very good profile.

Yes.

Uh, that the size approximately we're, we're working with usually 10, 20, 30 physicians.

Um, right now it's actually the very, I would say sophisticated, uh, expensive IT system for, uh, medical practices.

So we, we started with AI, but because of other needs and other requests, uh, right now, it's basically a.

I would say full blown digital front door, uh, system.

So you can, um, but in that other day, it's a communication device between patients and doctors.

So the AI, the, um, technology, the automation makes it that communication easier.

So in many cases, you don't have to call the office.

if you're patient, you don't have to sometimes to go there.

Of course, if it's something extreme, you know, obviously you, you, you do what you do, but the idea is, especially for more routine things, uh, routine symptoms or routine questions.

The organizations we work with, they are asking their patients, Hey, you know, you can go to the system 24 seven.

You can open this app on your iPhone.

Um, you can sign up for telehealth.

You can chat with a nurse.

You can, you know, access front desk.

You can schedule something.

You can access your actual medical records.

Without, without going to Epic.

Now there is a, there's an asterisk there that, you know, it, it really kind of depends, you know, what, um, you know, where you are, because if obviously if you have a doctor outside of our system, then obviously you're not going to have, you know, Have that, those records can, can only

see what the doctor has access to.

I would think, well,

whatever that practice has access to, the patient has access to.

So, but, but patient can actually add their, their own medical records.

So at the advantage is that they can quickly access our AI summaries and we have the tool called, um.

the concept cloud, which is basically in one visualization, you can see all of your problems.

Sort of, if you had something very persistent over the years, it would say something like, you know, uh, diabetes or something.

And, and it would actually, when you put a cursor on it, it would actually tell you, uh, what's your dosage, for example, for medication, for the particular problem.

So it's, uh, it's a lot of sort of, you know, automation, simplification using AI, but also using our other technology.