Episode 66: Building AI Apps for Healthcare and Biotech with Key Opinion Leader Jim St. Clair

Got an idea for an app? Well, in this episode, you can learn all about the Health Universe platform and how it can help you bring your idea for an AI healthcare app to life. You're listening to Bio from the Bayou, empowering biotech leaders with the tools, assets, and expertise to improve healthcare outcomes.

Each episode provides insights and information to expand your network, advance your product line, and help your company thrive.

Serving the entire Gulf South, but recorded in the heart of New Orleans. Welcome back to another episode of BIO from the Bayou, actually recorded at BIO on the Bayou. Today's guest is Jim St. Clair from Health Universe, a platform for building, deploying, and sharing health AI apps, and frankly, a Gulf South neighbor right next door in Mississippi. So welcome to the show, Jim.

Oh, thank you. It's great that you have me.

So give us a breakdown of Health Universe. What is it? What does it do? How can it help clinicians Absolutely.

We are a startup ourselves. It offers an online platform to be able to collaborate and build AI applications that we keep a running dashboard and make it widely available to anyone to be able to access and be able to use those applications.

Wow.

My particular focus is taking it into a direction with cause-based nonprofits— Dana-Farber, American Cancer Institute, American Heart Association— and helping show them the applicability of AI applications in their environment with their, uh, patient data. constituents and patients and advocates. We bring together then university and other developers so that we, leveraging Health Universe as a platform as a service, provide a platform for them to conduct hackathons and codeathons and develop AI apps that fit their use cases.

OK, great. Well, as a non— I'm a scientist, but I'm not on the technical side as it relates to AI. So what is the process of creating an app? How easy can someone like me do it?

Absolutely. Matter of fact, we have tutorials available. It especially uses something called Streamlit. applications, which are like a variation of Python, Python being the machine learning programming language.

Not the snake, nope.

Yeah, not the snake, though you can get tangled up in it. The platform allows you to build these apps custom designed yourself very simply, very easily, can be used and deployed shortly thereafter, after creating.

That's great. Wow. So as I understand it, last year, Health Universe hosted an AI hackathon. Can you walk us through the results of that and some of the apps that came out of there?

It was kind of the first proof of concept. It gave us an opportunity to demonstrate the viability of the platform. One of the big things that has come out of it is actually an open-source continuous glucose monitoring solution that our founder actually wears. So it really shows that you can pull together. I'm sure you had a similar experience. I know I did. I was caught up in the MIT hackathon during COVID and there was 2 or 3 others, and you're pulling across people and, hey, what great ideas can we have? Yeah, we've built upon kind of that fighting spirit. to organize this specifically for healthcare. There are other platforms that you can find on the internet commercially available to just conduct your own hackathon.

Yeah.

But we really want to do something that's focused on healthcare and cause-based applications and focused on building in the direction of AI and using AI code applications for things.

You kind of had a hackathon for your own hackathon.

Yeah.

Yeah. So were there any other— I mean, the glucose monitoring was one that came out. What was one that you're most excited about in general that has come out of the app?

You know, that's a great question. So I have been kind of leading the charge around doing this with a cause-based Yeah. I'm really excited to have started making contact with very large organizations and even the smaller foundations who are focused on, you know, challenging patient populations, n equals 10,000 maybe across the country for their specific conditions and just helping them understand, hey, all this fuss about AI, there's applications out there that maybe can help them with something on their phone for symptom tracking or the ability to gather more data to be able to understand their conditions on a day-to-day basis. And these are things that we know that there are sponsorship and benefactors for, working to get more attention brought to their specific constituents and, and what that cause-based nonprofit is focused on, giving the opportunity for students to participate in something meaningful as part of an exercise or their background, and then yielding some real results in the way of an app that maybe they want to keep it, they want to use it. Maybe it's something that gets commercialized. Maybe we just leave it to the open source community.

It bridges a gap because there's people on the healthcare or the science side that understand that piece very specifically. And then there are those on the coding side and the AI side, and they understand that piece. But there's such a— there's a divide of how do you take the data that these clinicians have been gathering over the years as they try to treat patients for whatever disease indication, and be able to take that data and do something meaningful with it.

Right.

It's one, it's hard in general, and you need the data scientists to help with that. But it's also from a non-AI technical person, It's overwhelming and you don't even know where to begin. So I love that you have something that can potentially help bridge that gap, that helps sort of cross that divide and, and make it a little more easily accessible.

That's absolutely the case. And, and I think you've characterized it very well. You have a population with a need and they know the use case, they know what they wanna see out of it, but they're not empowered to be able to, or adequately equipped to be able to understand, well, could I develop an app? Could I use some tool or something like that? Yeah. Meanwhile, of course, you have thousands and thousands of Well-intended students eager to learn, eager to apply their skills, that a focused use case that helps them work together as a team, design something new, yield positive results is just something they're looking for.

So, Jim, with AI technology evolving rapidly, how does Health Universe ensure that its AI-driven solutions are both innovative but also safe for patient care?

I love this topic. Matter of fact, I have a— I have a website.

It's a hard one. So good. I'm glad you do.

It is. And I have— I have a slightly differing opinion, only slightly. Well, I would say I have a slightly differing opinion that I have had more and more colleagues reinforce through their comments, ones that I've actually learned from other doctors who have said this, which is we get so caught up in AI and patient safety.

Yeah.

But yet we demand metrics out of AI in this early stage that aren't easily replicable from human care.

Yeah.

To quote an MD, rather than just take this from myself, he always says, compared to what? So we say, hey, AI is dangerous. Compared to what? Compared to a resident that's been working for the last 36 hours on 2 shifts and 3 shifts? Compared to somebody that's just having a bad day compared to the quality of the data that's in the EHR. I think what it is is that safety is very important, of course, but we can't view it through the lens that, well, we have to subject it to levels of consideration that the average patient engagement doesn't necessarily get subjected to. And that's the challenging part. There is kind of a flip side to it, though, which is depending on what AI sources are coming from or where they're being used or what vulnerabilities there are. I work quite a bit with the Coalition of Healthcare AI. And in our working group, as we talk about using generative AI, I try and bring up and suggest, well, are there considerations for using your own open source AI model locally for doing this instead of necessarily a partnership with a commercially available, and I won't name them, but the commercially available generative AI models. Those all go back to data strategy and data architecture, which is what is my, and it kind of goes back to our hackathons, which is what is my data and what do I want to do with it?

Right.

And what do I think answers can be? And how do I come about with approaching it and seeking expertise to decide what should I implement to make it work?

That's a great way of putting it as it relates to the safety issue compared to what— I mean, you think about rural medicine where there's not a lot of healthcare professionals in certain areas of the country and the world. What's worse is that they don't get healthcare or that they may get access to healthcare that is driven by some type of machine learning process.

God bless and thank you. And we can sit here on the Gulf South and talk about this.

All day long.

We come from rural states.

Absolutely.

And to quote where I am in Mississippi, The vast majority of counties have no OB-GYN services.

Oh, absolutely.

So I hear arguments about, well, it'll never replace docs, or of course, you have to consider how AI replaces docs. And I say, well, you got about 5 years. You have about 5 years between the gray tsunami, the rate of departure of clinicians, and not just because of burnout and the popular topics.

Which is important.

Which is important, yeah. But I think that accelerated the general demographic trend that the median age for the American Academy of Family Practitioners is 55. OK. 10 years, they're all retirement age. So what do we do then? That gives us 5 to 10 years to cogently and deliberately come up with strategies that AI steps in and either semi-autonomously or autonomously provides cares in ways that we haven't thought about yet. Because in 10 years, you won't have that human resource to go to.

Well, in a lot of ways, our industry has already been so safeguarded just by the very nature of what we do. We already have policies. I mean, if you develop a drug using AI, that's OK. You still have to go through the clinical trial process at the end of the day. Same thing with devices and a few other things of where we've built in a pretty good set of policies and rules. It doesn't matter necessarily how you got to that product per se. It still has to go through the same process anyway. So in a lot of ways, there's probably industries that could learn a little bit from the healthcare industry of how to set up sort of these processes.

Yeah, and I think on the flip side, it's also recognizing what our metrics for quality and delivery and performance now that can be weighed or judged against what you're doing for AI results.

Yeah, thinking about that now, not an afterthought, is always going to be key as it relates to AI. As AI continues to expand into the various aspects of healthcare, what are some other key areas where you see it has really the most potential for immediate impact?

So I'm also really excited about digital twins.

Oh, yeah.

Yeah.

So cool.

Oh, yeah, everybody lights up. It feels very space age, but I'm excited to see it.

It does, which always brings up the fact that the term came about in 1972 when they dumped the bucket of parts on the table and said, Apollo 13 only has this, make something work that we can use as a model. Absolutely.

Yeah.

And that was a digital twin. Part of that, just from a development standpoint, much of what we've done in biomedical research around digital twins is high-performance computing center and university researchers. But when we come back to AI and ML, what does it mean for me to have my own personal AI? Where does that data reside? How can I control that data? The confluence, I would say, between chip processing and chip power and the use of your phones and the use of cloud computing and edge computing Resources are, I think, all pointing us to a new paradigm for how you can have your own locally managed AI tools that represent that digital twin for you or for a condition that you have or for disease research.

I think one of the challenges of digital twins is that equity piece of it, of making sure that it's not just the rich people on the upper tiers that are able to afford their digital twin, but making it equitable and approachable. And I'm sure that having open source and having these sort of hackathons and making, lowering sort of the barriers to accessing this type of technologies. These are going to be really important piece of that equity part.

Yeah, I think that's brilliantly said.

And the more—

Great. I want, and now it's recorded that I, that I said something brilliant.

Brilliant.

It's on the record here at this point.

1:15. Yeah, that's right. I also do some work with the Mississippi with a nonprofit for coding academies, and we're trying to bolster the curriculum and availability of information, free courses on AI. And it's really just, the awareness to say, hey, these are tools that anyone can learn. And of course, coming from Mississippi or Louisiana, we always talk about we're trying to play catch up. And these are areas that we haven't excelled at. And I say, hey, AI offers you an asymmetric opportunity.

Yeah.

We can pull together a group of students from Jackson State University or from another small school and give them AI resources. And they can design and develop right alongside some crew from Silicon Valley with apps that are not only pertinent for all of the healthcare community, maybe they're specific to what they understand in a rural context.

Yeah.

That they know is important to their population.

The CEO of Genialysis, this AI pharmaceutical company, said that it's really sort of this gold rush era that I love and that it's really the tool users and makers that are going to be successful in this, this new sort of era of AI, of creating the tools to be able to use and mine the gold is going to be really key to success. So.

Yeah, absolutely. And again, I bring up data strategy, which is people are forced to recognize what data do I have and how do I want to use it?

Yeah.

Because the tools will be on their doorstep for them to do various things. And especially in a build versus buy world, it's, hey, I've seen 6 vendors today are offering me these AI solutions.

Yeah.

What are we gonna do about our data?

Yeah.

And figuring out from there good data governance practices and then deciding the right AI tools to use.

Well, we could talk about this forever. And in fact, we are about to go hop on a panel at BIO on the Bayou. So we probably need to save some of the little nuggets of gold. The juicy stuff. Yeah, the juicy stuff for when we're not being recorded and having it documented.

That's okay.

Not for attribution.

Absolutely.

It's still going to be recorded.

I just wanna put that out there.

Good.

It will be, I'll, I'll link it in the show notes. Good, good point. Good point. Thanks for reminding me of that.

But we'll have the corporate censors.

Yes. Yeah. That'll be for the cocktail party later. Well, are there any other upcoming events that you'd like to highlight where people can connect with you?

I'm always available on LinkedIn. I'm kind of one of those LinkedIn geeks that jumps on his messages. Yeah, absolutely. So look forward to connecting, look forward to messaging. And really, if anything I'd offer, there's such, as you all know, there's such a plethora of events coming up. Even now, in the course of the late part of the year, I mean, November conferences used to be verboten, and now I know of like five jam-packed.

Yeah, from it's it's a rush to the finish line here. So I can't believe it's almost November.

We need more conferences in August in New Orleans.

Please, yes. Well, I don't know. I don't know.

August in an outdoor venue—that's where the real people show up.

That's that's the true believers. Absolutely. Well, thank you so much for talking with me, Jim. We'll be sure to keep an eye out for the amazing apps that are coming out of Health Universe in the future. So as always, we will have Jim's contact information and informational links posted. It in our show notes. Make sure you check them out to learn more, and thank you once again for joining us.

Thank you. This has been wonderful.

Thanks for joining us for Bio from the Bayou, and we hope you'll join us again.

To never miss the latest updates in biotech, hit the subscribe button now.

We'll catch you on our next episode of Bio from the Bayou.