Breaking Through Healthcare's Most Challenging Frontier with Bobby Reddy Jr.

Khullani Abdullahi (00:01.218) Good afternoon and welcome to the AI in Chicago podcast. As always, I'm your host, Khullani Abdullahi, the founder of Techné AI, a Chicago based AI governance, risk compliance and strategy firm. As you know, AI in Chicago highlights operators, builders and thinkers who are scaling applied AI from the Midwest with implications for the entire world. Each episode delivers practical stories and actionable insights, empowering leaders, to use AI or understand AI minus all of the hype. Today, I'm pleased to welcome to the AI in Chicago podcast, Bobby Reddy Jr. Bobby is an actual practitioner of building AI systems and then implementing them and then commercializing them. He is the CEO and co-founder of Pernosys, a leading health tech medical device and diagnostics company in Chicago. Bobby holds a PhD in electrical engineering from the University of Illinois at Urbana-Champaign, where he contributed pioneering work in biosensors and nanotechnology for medical diagnostics. As a researcher and entrepreneur, Bobby and his company are shaping the future of sepsis acute care diagnostics in the US. Prognosis is at the forefront of AI-powered clinical decision support for sepsis. Under his leadership, prognosis has built the Immunix Precision Medicine platform where the company has attained FDA approval for its sepsis immunoscore product, which is the first and only FDA approved AI tool for sepsis diagnostics. The result is sepsis immunoscore, which helps clinicians save lives and reduce healthcare costs nationwide. Bobby, welcome to the show. Bobby Reddy (01:48.206) Thank you so much. Appreciate it. Khullani Abdullahi (01:49.996) Of course. As a CEO of a biotech company in the AI space and diagnostic space, I can only imagine the complexity of building prognosis to the point you have. Can you share with our audience a little bit about your background and journey and transition from academic research in nanotechnology and biosensors and then founding prognosis as a company and getting it all the way through? the FDA approval process. Bobby Reddy (02:22.094) Yeah, it's been quite a journey over the past 10 years. You know, I've always known that I want to start a company. But I didn't know what it was for. I just knew I wanted to start a company. So back when I was a postdoc at the University of Illinois in Urbana-Champaign, we were working on a technology to measure biomarkers from blood using a point of care device. And we thought that our technology was super cool, and that everyone was going to want it. Khullani Abdullahi (02:28.621) Okay. Right. Khullani Abdullahi (02:42.423) Okay. Khullani Abdullahi (02:46.966) Right. Bobby Reddy (02:48.528) It took us several years to realize that nobody actually wanted what we were building and I had to flip the script a little bit, you know, from a nail looking for a hammer to, you know, focusing on the applications. So one thing that really switched things around for me was that the hospital in town, Carl Foundation Hospital, actually gave me a card that let me get into different areas of the hospital. Khullani Abdullahi (02:59.959) Right. Khullani Abdullahi (03:15.086) Mm. Bobby Reddy (03:15.088) we were renting a room from them for the company. And part of that came as access to various different places in the hospital. I don't think they realized it at the time, but that card actually gave me access to everything in the hospital. It allowed me to get into patient rooms, it allowed me to get into the ICU, into the lab and everything. So I took full advantage of that. And for about two years, I spent a lot of time talking to doctors, patients, nurses, lab technicians. Khullani Abdullahi (03:28.557) wow. Wow. Wow. Bobby Reddy (03:44.943) all the different people in the hospital. And that's where things really started to flip for me from just wanting to build a company because I wanted to build a company to really focus on focusing on solving a huge healthcare problem. And that healthcare problem that we realized after talking to so many different hospitalists and patients, et cetera, was that we really don't understand the biology of patients in the acute care setting. Khullani Abdullahi (03:59.95) Right. Bobby Reddy (04:12.208) We've never run like a human genome project to understand patients in the acute care setting. And because of that, we just don't know how to treat these patients. And, we'll talk about it today, but for the past 30 years, there really hasn't been a lot of developments in terms of new treatments and new diagnostics for the acute care space. so, prognosis became fundamentally about, number one, enabling a deeper understanding of the biology. Khullani Abdullahi (04:12.248) Mm-hmm. Khullani Abdullahi (04:22.262) Right. Khullani Abdullahi (04:33.006) Right. Bobby Reddy (04:40.194) of patients in the acute care space, but then two, developing tools so that we can match the right treatment at the right time for the right patient. Khullani Abdullahi (04:49.23) Can you say a little bit more about what is unique about the acute care setting? Because I suspect there's something that has prevented innovation in that space or new diagnostics in that space. Why don't we understand the biology of patients in acute care settings? What would we need to do to be able to better understand that? Bobby Reddy (05:13.592) Yeah, it's multifactorial. It's pretty complicated, as you can imagine. But I would break it down into a few different reasons. What the first and foremost, the most important reason is that the biology is so complicated because these are patients that come in, they typically have a lot of comorbidities. know, typically when people come to the ED and they get hospitalized, they put in the ICU, a lot of them have complications. They have comorbidities. And so the biology is very complicated. Khullani Abdullahi (05:17.55) Right. Khullani Abdullahi (05:39.96) Bye. Bobby Reddy (05:43.759) It's not, you know, I don't want to say cancer is simple, but it's not as simple as a relatively healthy person develops cancer and cancer is the only thing that's wrong with them. And you just have to treat the cancer. Like these patients can have acute heart failure and have kidney injury all at the same time. They could come in with COVID, you know? So, so there's a lot of heterogeneity and complexity to the biology, which makes it just difficult to understand. second reason is that financially there's not a lot of incentive. Khullani Abdullahi (05:51.05) Wait. Wait. Khullani Abdullahi (06:04.078) What? Bobby Reddy (06:13.69) to for companies to develop innovative new products. This has to do with how the reimbursement is done for hospitalized conditions. Most reimbursement follows what is called the DRG, diagnosis related groups, reimbursement stratification. So basically that means that hospitals get paid the same amount regardless of the technology that they use or for a given condition. This is very different from outside the hospital where specific technologies can be reimbursed. Khullani Abdullahi (06:20.332) Okay. Khullani Abdullahi (06:26.061) Bye. Khullani Abdullahi (06:37.131) Right. Bobby Reddy (06:43.844) And so because of that, new companies don't necessarily want to develop innovative groundbreaking technologies because they're not going to get reimbursed for it. And so hospitals have to pay out of pocket. So that system has unfortunately really limited the amount of companies willing to go after the acute care space. Finally, I think that big pharma is just not interested in acute care. I've talked to so many people from big pharma and acute care is kind of an afterthought for them. It's viewed as a very small market. Khullani Abdullahi (06:49.738) Right. Khullani Abdullahi (07:08.355) Mm-hmm. Bobby Reddy (07:13.06) in a very difficult market to get into. Even though we spend 10 times the amount of dollars of all of oncology combined together on acute care. So if you add up all of the cancers combined together, we spend 10 times the dollars on acute care, which makes sense, right? Because these are the sickest of the sickest patients. They keep coming back over and over again. This is literally where all the healthcare dollars are going. And yet most of the innovation hasn't focused here, largely because I think big pharma is not interested in developing. Khullani Abdullahi (07:13.154) Okay. Khullani Abdullahi (07:22.936) That's insane. Khullani Abdullahi (07:30.996) Great. Khullani Abdullahi (07:36.088) Right. Bobby Reddy (07:42.768) new drugs, new diagnostics for the space. And that's largely because of the heterogeneity problem combined with the financial incentives problem. And all trials have basically failed over the last 30 years. And so Big Pharma kind of used the space as too difficult, too small of a market to get into. Khullani Abdullahi (07:48.396) Yeah. Khullani Abdullahi (08:03.382) I mean, I think this reminds me of Charlie Munger's famous adage, which is, show me the incentives and I will show you the outcome. And I'm really, I love the way that you categorize the challenges to innovation in that acute care space. Because I think that typology is, it's going to require policy reckoning, right? We cannot. Bobby Reddy (08:11.951) Yeah, Khullani Abdullahi (08:30.764) We have to alter the incentives in this space if we're serious about addressing that large amount of spend and quality live years lost from incidents in the acute care setting. Kind of drill, Bobby Reddy (08:44.023) And it's only getting worse actually because our population is aging. So I don't know, hopefully you haven't had to go to the emergency department recently, but the lines at the emergency department are getting out of control. And it's because all the hospital beds are filled and they can't admit people from the ED. So nowadays the ED is like a four to six hour wait time to even be seen by providers. So it's just getting worse and worse. Khullani Abdullahi (08:55.512) Yeah. Khullani Abdullahi (09:00.544) Yes. Khullani Abdullahi (09:06.008) Yes, yes, it is. it's, we're not aligning. our incentives to address, I think, both need from a patient standpoint and then from a care and innovation standpoint. One thing I will share with you is I had the privilege of working for an AI diagnostics company, not in the substance space, but cardiovascular disease. And I know from my work there that congestive heart failure is the single most expensive condition for hospitals in America. And similar to subsist, it kind of gets a lot of short shrift, although those innovations are beginning to change. So can you place first for my audience, which probably does not have a medical or biotech background, can you explain subsist, why you selected that as the disease state to build prognosis around? What about that disease state? And then the technology you were developing caused this like optimal combination and then talk about why subsist within the broader context of healthcare. Like where's the spend, where's the market, where's the need? Let's dive into that. Bobby Reddy (10:22.499) Yeah, absolutely. So sepsis is basically when you get an infection that leads to your body getting out of control, essentially. You know, we work on sepsis, but we also work on several other acute care conditions. So our first FDA approved product is in sepsis, but we're also working on acute heart failure, pneumonia, acute kidney injury, and a few other conditions. But focusing on sepsis for a second, sepsis is the leading cause of death in hospitals. Khullani Abdullahi (10:29.314) Yeah. Khullani Abdullahi (10:32.887) Bye. Khullani Abdullahi (10:42.574) I love that. Khullani Abdullahi (10:51.246) Bobby Reddy (10:52.431) It is one of the most expensive conditions that you can encounter in hospitals and it is a loss leader for hospitals, meaning that every single hospital across the United States is losing money on sepsis. And so because of that, it's quite an urgent medical condition to address. On top of that, the incidence of sepsis is increasing every year. Again, as our population is aging. Khullani Abdullahi (11:01.73) Okay. Bobby Reddy (11:19.311) As infections become more more common, as superbugs become more common, we're really headed towards a disaster in terms of antibiotic resistance, which is directly related to incidence of sepsis. So it's just getting worse over time. And fundamentally, there hasn't been a drug for sepsis in the last 25 years. Maybe 25 years ago, a drug called Zygres was approved by the FDA. Khullani Abdullahi (11:48.482) Okay. Bobby Reddy (11:49.456) It was Eli Lilly, the company that sponsored that, but the drug kind of was iffy. It kind of worked. It kind of didn't work. And it eventually got pulled due to safety reasons. So it's no longer available for use. So there's no drug for sepsis, no treatment for sepsis until our product was approved. There was no AI tool to diagnose sepsis. Sepsis is such a difficult condition because it's not just one signature. It's not just like, you know, breast cancer. Khullani Abdullahi (12:00.021) wow. Khullani Abdullahi (12:16.683) Right. Right. Bobby Reddy (12:19.339) it's there's 20 different sub conditions that are grouped under what sepsis is. Because sepsis more generally is just when your body gets out of control due to an infection, but bodies can get out of control in very different ways due to infection. The pandemic was actually a perfect example of this, where you and I could have the exact same infection, know, SARS-CoV-2, but I could be in the ICU and you could be perfectly fine. That just showed that we can have very different immune responses Khullani Abdullahi (12:40.855) Yep. Khullani Abdullahi (12:44.372) Right. Yeah. Bobby Reddy (12:48.505) to the same infection. And that is partially what makes sepsis so difficult is that there's so many different types of immune responses to a given infection. Khullani Abdullahi (12:57.208) So this reminds me of, is it the cytokine storms? Right? Right. Bobby Reddy (13:02.787) Yeah, cytokine storms is part of it, but there's something called a compensatory immune reaction that shuts down the cytokine storm. So you can actually become immunosuppressed after sepsis, and that's actually where a lot of the deaths occur. So there's a lot of complexity. Cytokine release storms is a big part of the initial reaction. On the back end, can have other reactions. people could have inappropriate cytokine release storms. They could have Khullani Abdullahi (13:07.714) Okay. Khullani Abdullahi (13:11.808) Okay. man. Khullani Abdullahi (13:19.822) Right. Yep, but then on the back end. Bobby Reddy (13:32.855) not aggressive enough or too aggressive. So there's all sorts of different combinations that we need to assess. Khullani Abdullahi (13:34.702) Right. It runs a continuum. So one, the reason I clapped earlier when you were telling me about your pneumonia in particular, expansion into pneumonia in acute care setting is because a lot of the emerging literature shows that you could have like the same doctors listen to pneumonia and the variance in diagnosing severe pneumonia is really, Bobby Reddy (13:57.443) Yeah. Khullani Abdullahi (14:04.512) is scary. So earlier you asked me, do you go to the emergency room? I have a seven year old and unless we really need to go, we avoid hospitals because I'm aware of the data on hospital acquired infections and misdiagnoses. and so there is a sense in which the technology that you're building allows us to begin thinking about supplementing, augmenting, extending. Bobby Reddy (14:12.335) Yeah. Bobby Reddy (14:17.335) Yeah. Right. Khullani Abdullahi (14:33.214) the accuracy of diagnostics. So I want to take a step back a little bit and think about AI in the medical diagnostic space from a research standpoint, right? From a implementation standpoint, from a physician preference feedback standpoint, just kind of Let's talk about that continuum because we are, it is the age of AI. is top of mind for all leaders. You have been working on AI diagnostics for far longer than just a handful of people in the world. What do people need to know about the role of AI in clinical decision-making for? aiding in diagnosing conditions like sepsis that are multifactorial, that are particularly complex, and where the data shows physicians struggled in accurately diagnosing it. Bobby Reddy (15:33.658) Yeah, I mean, we could start with your final point there, which is that there's a lot of literature that suggests that the doctors do struggle identifying sepsis. And it ranges in the full spectrum from a quote unquote good doctor to a bad doctor, but all the lies in between. But essentially, we've done a lot of work where we've asked the same doctor, sorry, three different doctors to look at the same patient's charts and to tell us whether or not the patient had sepsis. And this is Khullani Abdullahi (15:41.422) Mm-hmm. Khullani Abdullahi (15:47.596) Mm-hmm. Khullani Abdullahi (16:00.034) Mm-hmm. Mm-hmm. Mm-hmm. Mm-hmm. Mm-hmm. Mm-hmm. Mm-hmm. Bobby Reddy (16:03.235) like all of their charts, including whatever happened to them eventually. So they have the benefit of seeing the future outcomes of the patient. And more than 50 % of the time, the three doctors don't agree with each other. It's crazy. They're looking at exactly the same chart and they don't agree on sepsis. Yeah, yeah, Exactly. So it's kind of a mess in terms of diagnosing sepsis. Khullani Abdullahi (16:15.96) Jesus. including outcome data, like whether the patient died or lived or recovered. Bobby Reddy (16:32.897) And there's really no such thing as you have sepsis or not. It's a full spectrum. There's a lot of heterogeneity there. So, you when we developed our tool, we really focused on holistically evaluating the patient. And that's where the AI came to play. know, AI for us is just a tool. not, was not never the end game. You we were doing AI before AI was cool. But it was just a tool basically to assemble all of the data together. So. Khullani Abdullahi (16:36.971) Right. Khullani Abdullahi (16:51.662) Right. Right. Right. Khullani Abdullahi (17:01.387) Right. Bobby Reddy (17:02.259) The sepsis immunoscore uses vitals, it uses cell counts, it uses chemistry parameters, it uses demographic information, and it uses extra biobiological measurements from the patient. And it puts all of that information together to very holistically evaluate the risk of sepsis for aiding a physician. You mentioned a couple of times that this is really about aiding a physician, and that is really our messaging around the sepsis immunoscore. Khullani Abdullahi (17:29.345) Right. Bobby Reddy (17:31.023) We're not trying to replace anybody. This is not a standalone solution. This is like a calculator, you know, using instead of doing sums on a piece of paper with a pencil, you can use a calculator which makes it faster, makes it more accurate, makes it more usable. And so that's really how we're trying to use AI. In fact, I like to talk about augmented intelligence. It's resulting in augmented intelligence instead of artificial intelligence. Khullani Abdullahi (17:32.845) Right. Khullani Abdullahi (17:40.962) the Khullani Abdullahi (17:53.826) Right. Right. Yeah. And I think something that I think a lot about in the healthcare space where I spend a significant amount of time is helping leaders in healthcare move up the value chain, right? And so if I can accelerate the rate at which I can identify the problems and the comprehensiveness with which I can understand the patient's problems, then I can turn my attention to treating it and solving those problems, which ideally then also improves Bobby Reddy (18:26.38) Exactly. Khullani Abdullahi (18:28.46) I want to ask a little bit about, so I am very intentional about making sure we don't talk about AI systems that I can't explain. So I would like a mental model of the way this immunoscore works and take me from like research, right? So like, did you use blood-based biomarkers? Are you just, is it epigenetics and... Bobby Reddy (18:42.255) Okay. Khullani Abdullahi (18:56.504) genetics and methylation, are you looking at protein, right? Like walk me through how you collected and researched and experimented everything from like the biomarker stage and then on the AI side of things, how are you developing that clinical decision algorithm? Are you using an ensemble model? And then from a deployment standpoint, are you layering onto the EHR? Do you have a standalone device? Give us a concrete understanding of AI diagnostics in sepsis as the first FDA approved tool. Bobby Reddy (19:34.256) Yeah, so it all started with a data collection exercise and a sample collection exercise. I mentioned earlier that we don't have a human genome project for acute care, and it's going to sound a little grandiose, but we kind of set to do a mini version of that. So we collect, we, for over the last 10 years, we've been collecting a lot of blood samples from patients in the acute care setting. So it's now over 135,000 blood samples from over 32,000 patients. Khullani Abdullahi (19:38.238) Okay. Okay. Yes. Khullani Abdullahi (19:47.734) Yeah. I love it. Khullani Abdullahi (19:55.853) Okay. Khullani Abdullahi (20:01.398) Wow. Bobby Reddy (20:03.543) And it's constantly growing, you know, as we work with more more hospitals, more and more samples come in, more and more data comes in. And so, you know, we get all of the data about the patients that are measured in the hospital. So that's things like, you know, standard labs, vitals, medications administered, demographic information. So we get all of that. But what we do is we add on to that database with additional data that we measure from the blood samples. So we actually have what? Yeah, we actually have wet lab space. Khullani Abdullahi (20:15.242) Okay. Khullani Abdullahi (20:29.3) Okay, so you analyze it. Bobby Reddy (20:33.879) here in Chicago, where we literally take blood sample by blood sample out of the freezer, and then we thaw it and we measure a lot of extra biological parameters. They're called biomarkers that we measure from the samples. And this is stuff that they don't measure in the hospital. So this is how we get a deeper insight into what's going on with the patient. So we measure a lot of different biomarkers, and then we combine that data with the data from the hospital to get a very nice holistic view. Khullani Abdullahi (20:40.044) Mm-hmm. Khullani Abdullahi (20:45.966) Right. Khullani Abdullahi (20:50.657) Right. Khullani Abdullahi (20:54.84) bright. Bobby Reddy (21:02.991) of the patient. And you know, we're the only company that's done something like this, you know, the second closest initiative is probably a factor of 20 to 30 times smaller than what we've done. So we really have understanding of the biological insights into acute care patients that no one has discovered before. So that's really the research of where it all started. We use a bunch of different biomarkers, we use a bunch of different clinical parameters, and then Khullani Abdullahi (21:07.051) Right. Khullani Abdullahi (21:12.302) Yeah. Khullani Abdullahi (21:19.916) great. Khullani Abdullahi (21:23.79) Right. Bobby Reddy (21:31.856) we started to build different algorithms that use different combinations of these parameters. know, there's like 200 to 300 parameters you could use. We tried lots of different combinations. It was a combination of what was the most accurate for diagnosing sepsis, along with what is more convenient to measure in the hospital, along with, you know, there's this aspect of generalizability when it comes to AI algorithms where you could train an algorithm and they could work at 10 sites. Khullani Abdullahi (21:36.981) Right. Yes. Khullani Abdullahi (21:45.378) Right. Khullani Abdullahi (21:55.318) Mm-hmm. Bobby Reddy (22:00.867) but then you take it to an additional five sites and all of sudden it breaks. So when we were developing the album, yeah, sorry. Khullani Abdullahi (22:04.32) Right. Now you have... No, I was just going to say now you have model drift issues and now you have to go diagnose them and course correct. Bobby Reddy (22:15.853) Yeah, well, so model drift is slightly different from generalizability, but it's related. Generalizability basically means that it works at a bunch of hospitals, but for some reason it doesn't work at a few other hospitals. Drift is more that over time the model is drifting. So both the concerns are very valid concerns. But so we went through a lot of effort to optimize the accuracy, the usability, and the generalizability. Khullani Abdullahi (22:20.32) Okay. Khullani Abdullahi (22:27.798) Interesting. Khullani Abdullahi (22:42.305) Mm-hmm. Bobby Reddy (22:45.655) of the algorithm. And that actually was several years in the making, including up until the FDA approval, actually, the FDA process, which I think we're going to talk about in a little bit, was very helpful actually in making a better algorithm. The FDA pointed out a couple of parameters that were kind of iffy when it comes to the generalizability of the algorithm. So we actually removed those parameters and refined the algorithm as part of the FDA process. Khullani Abdullahi (23:01.582) That's why. Khullani Abdullahi (23:13.622) When you, so you started with, and I think AI does this really well, it helped you narrow down the universe of 200 parameters and biomarkers and data that would be relevant for accurate diagnostics. What did you end up with? Bobby Reddy (23:29.817) We ended up with 22 different parameters. So there's a few vital measurements. There's a few different cell counts. There's chemistry parameters. There's a demographic parameter. And then there's two additional biomarkers. So it kind of covers the whole gamut of holistic evaluation for the patient. Khullani Abdullahi (23:32.011) Okay. Khullani Abdullahi (23:43.191) Exa. Khullani Abdullahi (23:49.634) To give the audience an understanding of how long this process takes and the areas of expertise that were required, how long would you say it took from the sample collection and analysis algorithm development to FDA approval? Was that 10 years, 15 years, and what expertise? Bobby Reddy (24:11.277) Yeah, it was about 10 years from when we started collecting samples. So we started collecting samples and data even before we started the company as part of a postdoc at the University of Illinois. So it was over 10 years, honestly, of collecting samples, probably about five years to build the algorithm and to validate the algorithm. We got FDA approval in May of last year or April of last year, April of 2024. Khullani Abdullahi (24:15.596) Okay. Khullani Abdullahi (24:22.082) Mm-hmm. Khullani Abdullahi (24:37.794) Mm-hmm. Mm-hmm. Bobby Reddy (24:41.075) but we, we were ready with the algorithm at least two to three years before that. But we made a judgment call that we didn't want to start to sell until we got FDA approval. At the time, it wasn't clear whether or not the FDA needed to regulate because it's so new. It wasn't clear whether the FDA needed to regulate the space or not. But we kind of drew a line in the sand and we said, this definitely affects medical decisions. These are life or death situations. And so we want to, we want to get FDA approval. When I'm still. Khullani Abdullahi (24:46.552) Right. Khullani Abdullahi (24:51.182) Pray. Khullani Abdullahi (24:55.214) Right. Khullani Abdullahi (25:06.637) Right. Bobby Reddy (25:10.627) Confused that we are the first ones that got FDA approval because there were so many people working on algorithms, but no one has gotten FDA approval until we did. And we remain the only people that have FDA approval. Khullani Abdullahi (25:15.841) Right. Khullani Abdullahi (25:19.309) Yes. I think that as someone who's worked with companies who are going through that process or have gone through that process, it's painful and it is extensive and it is significant. But what makes it easier is when you have companies built by sound researchers and sound research, right? And so walk me through kind of that. Bobby Reddy (25:44.46) Right. Khullani Abdullahi (25:48.384) So I think like you knew you were, you wanted to be an entrepreneur, even as you were going into academia and getting a PhD. Did you like walk me through the business journey and business pivot? If you always intended to kind of start a prop up. company, was it easier for you to pivot from a researcher into the role of a CEO now dealing with investors and the media and competitors and potential venture partners, et cetera? I mean, that is a big jump. And then that process kind of feeds into the FDA being like this key stakeholder to validate the value and contributions of the company. Bobby Reddy (26:31.661) Yeah, so it's been quite a journey for me personally. It started as kind of a concept that I wanted to start a company, but I had no idea. Anything that was involved from the business side, from the technical side, from regulatory side, to a lot of lessons that I've learned over time. One of the big lessons that I learned, as I mentioned early on, was that we needed to be problem focused. We need to be problem and mission focused, not business focused, not Khullani Abdullahi (26:35.371) Right. Khullani Abdullahi (26:41.292) Mm-hmm. Right. Khullani Abdullahi (26:48.558) Right. Khullani Abdullahi (26:56.321) Bye. Khullani Abdullahi (27:00.846) Bobby Reddy (27:01.807) Obviously the end goal of a business is to make money, but that can't be the sole goal of the company. The company needs to be driven by a mission and that mission needs to be grounded in a real problem that society faces. And so we found that, in retrospect, found that somewhat early, even though it took us several years to find that. We found that core problem that we want to solve, which is basically that people don't understand the biology of patients. Khullani Abdullahi (27:08.142) Yeah. Khullani Abdullahi (27:16.642) Great. Khullani Abdullahi (27:24.129) Right. Bobby Reddy (27:31.204) And there's no tools to in real time to evaluate the biology of those patients. So that was the problem that we set out to address for good or for bad. No, no company is really trying to address this problem. Even though it's such a big problem and it's such a big market because of variety of different reasons, there's not really another company that's trying to address this as of right now. So, so we, had a lot of time, you know, it wasn't like competition was breathing down our necks. Khullani Abdullahi (27:35.384) Right. Khullani Abdullahi (27:45.95) Right. Khullani Abdullahi (27:56.748) Right. Right. You're right. Bobby Reddy (27:59.524) we had a lot of time to think about things very carefully and to do things the right way. Since then, my role has evolved quite a lot. used to be in the lab measuring blood samples or computer developing algorithms. My role has evolved now to mostly dealing with investors, dealing with team dynamics, dealing with inspiring others to commit to the work. Khullani Abdullahi (28:07.886) Okay. Yeah. Bobby Reddy (28:27.631) So it's definitely evolved significantly since then. But it was kind of, you know, it wasn't an abrupt transition for me at least. You know, I know in other companies, you you go and you raise VC dollars and all of a sudden you go from, you know, not even having a real company to raising a $50 million round and you go and you hire 40 people and the transition is very abrupt. For us, we did things a little differently. We didn't raise big VC dollars. We kind of, Khullani Abdullahi (28:34.51) Right. Khullani Abdullahi (28:42.026) overnight. Khullani Abdullahi (28:50.251) Right. Bobby Reddy (28:56.579) We raised money slowly and we supplemented a lot of it with grant funding. So we've done over $30 million in grant and non-dilutive funding to complement the dilutive funding that investors have raised. And so for me, it was a little bit of a slower transition. We went from five people to 10 people to 11 people to 13 people, you know? And so it was kind of more of a natural transition. And I picked up the business stuff along the way. I'm still learning a lot about the business stuff. Khullani Abdullahi (28:58.318) Right Khullani Abdullahi (29:04.312) Wow. Khullani Abdullahi (29:16.994) Mm-hmm. Bobby Reddy (29:24.663) as we learned every day. But it was a little bit more of a natural slow transition for me instead of just being thrown into the mix like I know some CEOs are. Khullani Abdullahi (29:24.792) Bye. Khullani Abdullahi (29:37.314) that they experience. How has, so you worked on AI before it was a buzzword and before everyone was demanding a sound AI strategy from their leaders. How has the way that you talk about AI with hospitals and clinicians changed since you had that all access pass at that one hospital back in the day? Bobby Reddy (29:46.991) Okay. No. Bobby Reddy (30:03.649) Yeah, well, we talk about AI kind of as a secondary line of sales. mean, the first line is that you need help when it comes to Sepsis. And we built a tool that better helps you address and evaluate Sepsis. AI is just a piece of the tool. It's just a feature of the product. It's not the leading reason to adopt our product. So. Khullani Abdullahi (30:08.256) Okay. Okay. Right. Bobby Reddy (30:30.223) So yeah, AI just exists in order to combine all the parameters together in an effective fashion. Besides that, really the goal of the tool is to holistically evaluate the patient by taking into account many different types of parameters in a way that human beings can't memorize all these different parameters and memorize all the different trends of this being high and this being low and this being medium. So it's just a tool to help. Khullani Abdullahi (30:36.652) way. Khullani Abdullahi (30:42.956) Right. Khullani Abdullahi (30:48.558) All Right. Right. Bobby Reddy (30:56.097) evaluate and to do pattern recognition. That's really truly what it's doing. It's doing pattern recognition to recognize the patterns that could lead to deterioration. Khullani Abdullahi (31:06.542) What kinds of questions are hospitals and physicians asking you when you talk about AI-enabled sepsis diagnostics to them? Yes. Yes. Bobby Reddy (31:19.011) Yeah, so hospitals and clinicians are two very different audiences, both of which we have to convince. Clinicians tend to be very convinced by the data. So we have a New England Journal of Medicine AI paper that covered the FDA. Also, the FDA approval helps a lot to support our claims of accuracy. So clinicians tend to buy in after a couple of sessions. They're mostly interested in the accuracy of the tool. Khullani Abdullahi (31:29.098) Okay. Okay. Khullani Abdullahi (31:38.251) Right. Bobby Reddy (31:48.592) how much it can improve their decision making. We have a lot of data that shows that the sepsis immunoscore does improve the decision making of clinicians. It's much more accurate when you use the tool plus a clinician compared to a clinician alone. So when it comes to clinicians, it's really about accuracy, speed, ease of use. A big feature of our product is we actually display to the end user all the different parameters used to calculate the score. Khullani Abdullahi (31:48.597) Okay. Khullani Abdullahi (32:02.827) Okay. Bobby Reddy (32:17.325) on a patient by patient basis. That was really important because a lot of AI tools for sepsis and other conditions are very black boxy. You don't know what goes in, you only know what's spit out and you don't have no idea how it was calculated. So for us, big part of what we do is we literally show every one of the 22 parameters, which parameters increase the risk of sepsis, which parameters decrease the risk and by how much. Khullani Abdullahi (32:23.918) with bucks. Khullani Abdullahi (32:30.774) Yeah. Right. Khullani Abdullahi (32:43.009) Wow. Bobby Reddy (32:44.077) Because a big part of a big piece of what we do, I always say that that the sepsis aminoscore is great, but it's not always it's not perfect. It's not always going to get things right. And in particularly when it gets things wrong, it's very important for clinicians to forgive the tool, to be able to forgive the tool, to be able to see, these parameters were elevated. I see why the aminoscore thought that patient was going to be something miscephalic. But I know that actually it's something else that caused those parameters to be elevated. But I forgive the tool because I see why it made the mistake. Khullani Abdullahi (32:51.432) Right. Right. Khullani Abdullahi (32:59.219) Right. Right. Khullani Abdullahi (33:03.81) Bye. Khullani Abdullahi (33:08.75) Right. Khullani Abdullahi (33:13.282) Right. Bobby Reddy (33:13.411) So it's really important to have increased transparency about how the tool calculates things. So anyway, that's all kind of clinician facing. On the hospital side, hospitals are the tougher audience because ultimately they really care about ROI. They care about how much they're going to spend on the product and what is the benefit we're going to get from the product. So for them, we have two kind of main tracks of convincing. Khullani Abdullahi (33:18.029) Right. Khullani Abdullahi (33:26.314) Mm-hmm. Khullani Abdullahi (33:30.51) Right. Bobby Reddy (33:42.916) The first track is basically that we can improve their quality metrics. So every hospital in the United States has to report back on something called their step one bundle compliance, which is a measure of how well the hospital is dealing with sepsis. And the average right now is around 50 to 60 % compliance. So it's quite low. So most hospitals are very eager for anything that can make that number better. So the sepsis immune score directly can make that number better. And then the second piece is kind of the finances, the costs, the revenue that they generate minus the costs. So we can help both on the revenue side because a lot of sepsis is missed and it's missed in the billing even, so they forget to bill for it. So the sepsis immunosor can help to encourage them not to miss billing. So we can increase on the revenue side, but we can also decrease in the costs with improved care. So all of that goes into making the sale. It's a quite complex sale. Khullani Abdullahi (34:21.208) Bye. Khullani Abdullahi (34:37.324) Interesting. is. Well, the technology is complex. And we know that like adoption in provider organizations can be a little bit of a lag. Going back to the question of like implementation, are we when someone adopts the prognosis platform, our clinicians, do they draw blood and then send it to your lab for analysis or Is it just integrated into the EHR and that you just pull those parameters for analysis and you're able to kind of turn it around? What does that look like in terms of point of care? Bobby Reddy (35:17.923) Yeah, so it's fully accessible at the hospital, so they don't need to send the samples out to anyone. The way that it works is that if a provider suspects sepsis in a patient, they can order the sepsis immunoscore. Alternatively, most hospitals have early warning systems when it comes to sepsis, so it can also trigger off of an early warning system. The problem with early warning systems is they have very low positive predictive value, meaning that out of 100 times that the alert system fires, Khullani Abdullahi (35:23.137) Okay. Khullani Abdullahi (35:31.692) Okay. Khullani Abdullahi (35:35.202) Mm-hmm. Okay. Khullani Abdullahi (35:44.301) Mm-hmm. Bobby Reddy (35:47.652) Typically, it's only correct 13 to 15 times. So 85 times, it's incorrect. So the sepsis immunoscore can really help early warning systems be better because every time the early warning system fires, you can order the sepsis immunoscore. And what the sepsis immunoscore does is it looks for missing data. So the dirty secret about AI is it doesn't do well with missing data. And the sepsis... Khullani Abdullahi (35:51.544) That's very low. Yeah. Khullani Abdullahi (36:08.045) I'm right. Khullani Abdullahi (36:12.718) It just fabricates. Bobby Reddy (36:15.503) Yeah, it just, you know, most AI algorithms are just making up, they're imputing the missing data. The Cepstamuniscore, what we do is we actually force the hospital to order the missing data. So we don't say, hey, you're missing data, I'm just going to fill it in and pretend like I have the right value. Yeah. We say that we ask the hospital to order the missing data. So that includes all of the 22 parameters that are missing, but specifically it includes two biomarkers. Khullani Abdullahi (36:19.086) Yeah. Khullani Abdullahi (36:26.862) Mmm. Khullani Abdullahi (36:31.936) I know, right? Khullani Abdullahi (36:37.707) Interesting. Bobby Reddy (36:45.167) that they never really order for these patients. So that's pro-calcitonin C reactive protein. So we force the order of those two proteins in addition to 20 other parameters. And now when all of those parameters come back, now we can become much more holistic and much more accurate in diagnosing sepsis. So it's a really shift in thinking. Instead of AI algorithms just being okay with missing data, we took a stand and we said, we are anti-missing data. Khullani Abdullahi (36:47.369) Right, right. Khullani Abdullahi (37:05.036) Yeah. It's interesting. Yeah. Bobby Reddy (37:14.403) We want you to order all this stuff because it's really important to get this right for the patient and you need to have a trustworthy FDA approved result that up. Khullani Abdullahi (37:21.912) Right. It's fascinating because humans, we do this a lot where we, especially in the business world that I've spent time in, I've seen you need to be able to act with uncertainty. And it's important that we don't want our algorithms, right? Especially those that we put in positions of helping us. make life or death decisions, just be willy nilly free to make decisions in the absence of data. Humans do that all the time, right? Like we do that visually, we do that biologically. We just don't have a comprehensive stream of the real world as it is. We do a lot of filtering and we somehow manage to drive a car, make it to a home, et cetera. I love this notion of building an AI system and preventing it from, filling in the gaps as it were for data that it doesn't have, even if it may be able to do that relatively well and for triggering a data creation exercise that can then feed back, right? Because in business, we talk about building solutions to problems, but there's a data creation, there's a gap of unknown unknowns. So you've identified the unknown unknowns we need to close the loop on, which is these 22 parameters. And now you're able to say here, Bobby Reddy (38:20.931) Right, right. Bobby Reddy (38:30.06) Exactly. Khullani Abdullahi (38:49.36) the parameters that are missing, trigger an order for those tests, close the data creation loop, now analyze this more holistic patient's set of data, and then provide a enhanced framework for decision making. Can you share any data on what the improvement is? What is the positive predictive value of immunoscore compared to like a human? just a human clinician. Like, are you able to improve decision making 30%, 50 %? I have not read your FDA report yet, but I will. Bobby Reddy (39:26.337) Yeah, yeah. Well, the FDA report doesn't contain a comparison to clinicians. But just to give you an idea, know, early warning systems, the positive value, the maximum you see in literature is about 13 to 15%. And that's that's on a good day. The positive value of the sepsis immunoscore is more in the 65%. So it's a significant improvement over early warning systems for sure. When it comes to clinicians, it really depends on the patient. Khullani Abdullahi (39:31.403) Okay. Khullani Abdullahi (39:38.651) Mm-hmm. Mm-hmm. Khullani Abdullahi (39:45.3) Okay, that's significant. Bobby Reddy (39:54.682) But anywhere between 10 to 20%, it's more accurate than the clinician by themselves. Khullani Abdullahi (39:54.743) Mm-hmm. Khullani Abdullahi (39:58.67) That's it. And it's faster, right? Because it's going to allow clinicians to say, I'll just order the data that I need and make a decision at time two, rather than make a decision at time one that has implications. clinicians, so I think something that is in the backdrop. Bobby Reddy (40:11.191) Exactly. Khullani Abdullahi (40:21.21) is responsible AI, ethical use of AI, trustworthy AI, right? Like we're seeing all of that. Are you finding, are there certain clinicians that are comfortable with augmenting themselves with this augmented intelligence? What trends are you seeing in physician levels of comfort with adoption? I know, you know, there's generational issues, you know, I've been to the American Medical Cardiology Association's meetings and I've had cardiologists yell in my face that I cannot use epigenetics, genetics, and AI to better predict the risk of a heart attack. And I can, right? Like the literature is there, validated at Intermountain Healthcare. And so I've seen that side of it. What are you seeing for clinicians on your side? Bobby Reddy (40:58.414) Yeah. Bobby Reddy (41:07.255) Yeah. Yeah. Bobby Reddy (41:17.187) Well, I think that it's a little bit of a special scenario for us because most clinicians realize that they need help when it comes to diagnosing substance. It's not as easy as just, you you figure it out and you diagnose it. So we haven't gotten a lot of pushback from clinicians. Part of what we do also is we're not doing generative AI. We're not doing kind of the more fancy new age AI stuff that the most people associate with AI. Khullani Abdullahi (41:23.938) Mm-hmm. Khullani Abdullahi (41:30.229) Right. Khullani Abdullahi (41:38.03) Right. Right. Right. Bobby Reddy (41:44.312) our stuff is kind of more vanilla. It's more, it's random forest models. It's combining different types of data together and it's all quantitative data. It's not, it's not reading charts, you know, and, and, and, with all the caveats that comes with, so it's all measurements from the patient. And so because of that, I think it's much more objective than, you know, generative AI could be. So we really try to segment ourselves as we're basically a calculator. Khullani Abdullahi (41:51.886) Right. Khullani Abdullahi (41:58.72) Right. Khullani Abdullahi (42:06.208) Mm-hmm. Khullani Abdullahi (42:13.198) Right. Bobby Reddy (42:13.225) where something that can help you do sums much better than if you're doing it yourself. It's a very difficult diagnosis of sepsis. And so you can use any help that you can get. So we don't get too much pushback from clinicians. We usually, clinicians are very excited and they want to use the tool to improve care. Where we get pushback is more from the hospital. Because the hospital wants hard proof that this is going to affect their bottom line. Khullani Abdullahi (42:29.486) Right. Khullani Abdullahi (42:36.161) Right. Right. Bobby Reddy (42:41.903) and we're doing studies to show the impact on a hospital's bottom line, but just because you can impact one hospital's bottom line, then another hospital doesn't think you can, so it is challenging to convince hospitals that they should adopt. Khullani Abdullahi (42:48.066) Yeah. Khullani Abdullahi (42:52.557) Yeah. Khullani Abdullahi (42:57.024) and pairs, those real world evidence pilots are very painful. As we come up on this conversations and I do want to give you a chance to kind of reflect on how you see AI evolving in the acute care space, I'm hoping that prognosis. is a harbinger of other companies wanting to serve this space and innovate. But I'd love to see kind of how you're thinking about that evolution. Any advice that you would offer to other researchers or engineers who might be thinking of jumping into the entrepreneurial gauntlet. And then what's your favorite thing about Chicago? Restaurants, concerts, what have you. Are you Chicago native? And then we'll wrap up. Bobby Reddy (43:19.011) Yeah. Bobby Reddy (43:39.47) Yeah. Okay. So your first question was related to, sorry. yes. Yes. Well, what I really hope is going to happen is that prognosis or a combination of companies like prognosis are going to show that drug development is actually possible in the acute care space. So the way that we address drug development, we didn't talk too much about it today, but we basically use these AI algorithms Khullani Abdullahi (43:46.575) AI in the acute care space over the next few years. Bobby Reddy (44:09.049) to reveal the true biological state of a patient and to subdivide different patients into subgroups of biology so that each of them we can divide and conquer. We can develop drugs for each of the specific types of biology. Now you can't do that unless you have deep understanding of the biology, which is why we built our biobank and our data set. So what we really hope, and this is what we're setting up for now, is we're setting up clinical trials, a new type of clinical trial. Khullani Abdullahi (44:13.048) Wait. Bobby Reddy (44:37.113) where we'll measure the biology first and then decide whether or not the patient should be in a trial for drug X. And that patient might not be in the trial for drug X, but they could be in the trial for drug Y because their biology is different. And so what I really hope to show within the next three years is that this type of trial actually changes the landscape, meaning that this type of trial reduces the sample size needed of patients in the trial, but actually increases the signal and that for the first time we can have successful clinical trials. Khullani Abdullahi (44:43.469) Right. Khullani Abdullahi (44:47.523) Anyway. Khullani Abdullahi (44:57.282) Mm-hmm. Khullani Abdullahi (45:04.279) Mm-hmm. Bobby Reddy (45:07.129) with drugs in the space. And if we do that once or twice, I have a feeling that big pharma is going to pay attention because the market size is enormous. It's huge. Right. Khullani Abdullahi (45:17.228) Yeah, and you change the economics of it, right? So, so much of the money in clinical trials is wasted, but now you're able to increase my ability to filter on the right patients before I enroll them. That's extraordinarily exciting. Bobby Reddy (45:28.726) Exactly. Bobby Reddy (45:33.718) Exactly. So my hope is that, you know, three to five years from now, we'll have a like a craze like the GLP one craze happened with big pharma. I want to see a craze like that within the next three to five years in big pharma and a push towards developing a whole slew of new medicines for acute care. And whereas right now there's not much much activity, I want there to be a ton of activity. And obviously from kind of a selfish perspective. Khullani Abdullahi (45:40.568) Yes, yeah. Khullani Abdullahi (45:53.102) Right. Khullani Abdullahi (45:58.318) Right. Bobby Reddy (46:01.453) I hope that prognosis is the center of that because we'll be the only ones that can reveal the biology. so anyone that wants to develop drugs will have to go through us. But I realized that it's a bigger problem than just us. So other companies will eventually start to do the same thing. But that's my kind of longer term dream is that big pharma and bigger companies are going to really aggressively start to go after acute care because something needs to change. We can't keep doing things the way that we were doing. We need a paradigm shifting approach to change acute care. Khullani Abdullahi (46:04.246) Yeah. Right. Khullani Abdullahi (46:10.015) Right. Khullani Abdullahi (46:25.046) Right. Khullani Abdullahi (46:31.584) I love that. What advice do you have for researchers or engineers who might be considering bringing their innovations out of the lab to the marketplace? Bobby Reddy (46:36.932) Yeah. Bobby Reddy (46:41.015) I think the number one thing is something we talked about earlier, but it's solve a problem. Don't focus on building a company. Don't focus on building a product. Make sure that at the core of why you get up every single day is to solve a real problem. And it should be a significant problem that society faces. Because as long as you're focused on a real problem that society faces, everything else will fall away. You'll build the right product. You'll build the right team. Khullani Abdullahi (46:45.634) Right. Bobby Reddy (47:10.407) You'll raise money, you know, as long as it's a significant problem that you're focused on, but don't get caught up too much too early on the company or the product that you're building. I feel like so many of our engineers, our fellow engineers, or even just entrepreneurs, we get so caught up in, I have the best product, not that I'm solving a huge problem. Khullani Abdullahi (47:27.992) Yeah. Right. That someone's willing to pay me for in a timely fashion, right, and invest in. Are you a Chicago native? What do you love about Chicago? Why Chicago? Other than you went to school here and you're stuck here. Bobby Reddy (47:34.093) Yeah. Yes, yes, exactly. Exactly. Bobby Reddy (47:43.587) Yeah. Yeah. I'm not a Chicago native. I grew up in Southern California, but I've been in the Midwest now for more than 20 years. And I love the, I love the culture. love the people. I love the grounded nature of the people, as opposed to the coast where people can be a little more salesy. I feel like people here are more grounded. They're more real. They put their head down and they do real work without looking for all of the flashiness and all of that. So I really liked that about Chicago. Khullani Abdullahi (47:50.636) Mm-hmm. Khullani Abdullahi (47:56.534) Right. Bobby Reddy (48:11.279) I love the summers. We just officially ended the summer a couple days ago. Yeah. But I love the energy that the summer brings to. I grew up in Southern California where every day was perfect weather. so nobody ever appreciated it. But here I feel like when summer rolls around, everyone is in a good mood. It's just a crazy energy. Khullani Abdullahi (48:13.13) Yeah. We have three weeks left. Khullani Abdullahi (48:25.724) summer. Khullani Abdullahi (48:33.694) Yeah, it's very choious. As CEO, you do a fair amount of recruiting. Do you want to make a pitch to people who are looking for internships or roles or looking to invest? Bobby Reddy (48:46.285) Yeah, mean, we're, Prenos is working on really cool things. It's a huge market. We have an innovative approach. We're probably the only company that I know of that is working in the space. And we could have a massive impact if we're successful. So I encourage you to apply, invest, et cetera. Khullani Abdullahi (49:04.384) Excellent. Thank you so much for taking the time. AI in Chicago podcast community, you can find Bobby ready on LinkedIn. I will also link in the show notes, Pernos's website, exciting things coming from our own backyard. I always say Chicago is well positioned not only to win in quantum, but also to win in AI, especially at the intersection of biotech. And I'm pleased to have people like Bobby building here. Until next time. Thank you so much again, Bobby. Bobby Reddy (49:33.53) Thank you so much. Appreciate it.