Turning Healthcare AI from Vision to Verified Impact with Pegasus One

Welcome to the AI-Advive series on the BEAT Podcast, where we dive into the artificial intelligent solutions that are driving automation, generative, and agentic AI in the healthcare space. Follow this series to develop the skills you need to evaluate use cases, plan your next move in the world of democratized experimentation, and stay in the know on the latest AI technology breakthroughs. And now, your host, Sandy Vance. Hey, everybody. Welcome back to the AI series on the BEAT. I'm your host, Sandy Vance. And today, we're here with Pegasus 1, Tushar Piri is the CEO and founder of this services organization that has been around for about 15 years. Welcome to the show Tushar. Thank you, Sandy Vance. Great to be here. And we are delighted that today Tushar has brought one of his clients, Sebastian Uslas, who is the co-founder of Charter Health. Sandy, excited to be here with everybody? Sebastian is a technical co-founder of Charter, and they have been developing a platform technology that Tushar's organization has helped with. So we're going to get into a deep dive of how they're building the secure and scalable AI platform. But first, Tushar, can you just tell me a little bit about Pegasus 1 and where you started? Pegasus 1 is a healthcare focused technology and product engineering partner. But since the inception, I never considered us to be a company who is just providing services in building products because we realize and we realize it every day that for building a solid product, companies like us need to understand business as well as our customers too. And then look into the product development more from owners perspective. And that's what we bring to the table for each and every offer implementation, especially in healthcare, sadly, about 80% of AI initiatives do not make it to production. And we have seen the reasons most of these implementation do not make it to production are predictable. And they are generally ignored, such as aspects like interoperability or data latency, which are generally coming in as afterthought. And workflow and integrations are generally underestimated or ignored. So we have essentially codified these aspects which can make a project to not succeed in what we call as a song framework. So when we work with any in every client, not only that we focus on their core problem or the initiative they are trying to build, but also ensure from day one that they don't into this pitfalls. So we are building with the end in mind always and ensuring that their product is indeed in a position to go to market beyond POC. Tell me more about this song framework. Song framework stands for as for signal, which we will call the data aspect orchestration stands for the workflow aspect. Normalization is normalizing the data, you know, as you know, same medication is written differently by different doctors and facility. Similarly, lab results, even imaging results for same kind of scan are coded differently. So for a company like charty, all these aspects need to be normalized so we can get more coherent outcomes. And then G stands for governance, which as you know, it's obvious because for any and every implementation governance need to be baked in from day one. So when we look at a problem or a project, we ask these questions asked to ourselves and also to our clients, every thought through the data aspect, the orchestration, which is will it work in the workflow, which clinicians are already using or you're expecting to change their daily routine, which means it is better how good your AI is. It's not going to be very successful if you're asking clinicians who are already overburdened over work to change their day to day work. And normalization is obvious, you know, especially for project product like charty, we are working with humongous amount of data and we need to make sure if all the data pieces are normalized and they talk and mean same thing, then getting lost into translation of the various system, then governance has to be the part of the whole application. So we can identify ever changing ecosystem of healthcare and we know what's changing, why it's changing and we can explain it and continue to build the trust of our clinicians. Fantastic. I love a good acronym. I think that is such a cool way to sort of make sure that you hit all of the things. So Sebastian is the technical co-founder of Chardar as we discussed and he's been working on shaping the platform's technology vision and architecture. So tell our audience what is this application that you're trying to build or is it a platform you're trying to build, how you got involved with Pegasus 1 and how they've helped aid in your mission. The vision of Chardar Health overall is to be able to allow hospitals and then health systems to use their data across different data silos. So think unstructured data in your EMR, different data systems that they have, financial data and leverage that to kind of generate analytics and understanding from that. So an example that we like to use is lots of hospitals care a lot about their sepsis rates because CMS oftentimes penalizes hospitals based off of their sepsis rates and they try to really understand what's driving those sepsis rates to be bad and then really work at addressing those issues to be able to fix it to be able to kind of deliver higher quality of care for patients as well as reducing those fines from CMS. And a lot of the time the work that they do to do that is they're higher in consultants to really go in, look at the data, analyze and understand what's happening with that data. But those things can take quite a long time and it can take months or partially up to a year to be able to understand and analyze that data and then they have to go in and actually implement those changes. What we're working at at Chardar Health is a system to be able to allow these hospitals to continuously analyze and understand the information that they have across their different data silos. And what we're doing directly with Pegasus 1 is building out a product called Chardar, which is a autonomous system that allows you to take that data and ask a question similar to how you'd ask Chatchee BT or ask kind of cloud or things like that to be able to say from this data that we have about these different patients, what are some kind of questions that we'd like to ask and then generate those answers. So as an example, one of those things is saying for these different patients, what's their right of sepsis? And then we can ask even more complicated questions from there like what are some factors that correlate to these sepsis cases? As an example, maybe these patients were seen in a certain department of the hospital and that correlates quite heavily with these kind of increased rates of sepsis. And those types of things are things that people within the hospital, within their data departments, are doing already. And they're truly trying to understand that. And we're really working on giving them a system and a platform to do that automatically without having to basically take the data out of those silos and then put it into Excel and in all these different places, which can take quite a lot of time and they don't really have yet a unified system to be able to do this type of analytics. And that's really kind of the goal with what we're building. Fantastic. So what made you decide to partner with Pegasus? Yeah. So Pegasus one was one of the kind of best people that we found in terms of both their understanding of kind of healthcare interoperability standards in terms of kind of fire HL7 as well as kind of how those systems work in general. We as our team did not have as much of kind of a technical background on that side. Our expertise is more so on that kind of clinical data processing, processing unstructured tech side of things, as well as kind of the AI agent side of things and really kind of what we saw from Pegasus one was quite a lot of expertise when it came to interoperability as well as kind of how we can think about those kind of solutions that we were thinking about long term. As Tushara had mentioned, they really worked with us to understand, okay, what is it that you guys are trying to do? What is it we're trying to develop? And then is that feasible with those technologies that exist within the healthcare world? And we work directly with them to answer a lot of those questions that we had in terms of kind of applications of AI to these places. What's that data that we can get from the different smart on fire APIs and things like that? So they've been a great help when it came to kind of setting the direction for our technical side of our product and really help us understand what is feasible and what's kind of capable within those healthcare settings. Fantastic. So Tushara, you mentioned earlier that a lot of AI projects fail. Obviously Pegasus one works with organizations. You want them all to succeed. Do you sort of have to make decisions going into projects like Sebastian presented to you to figure out if they're going to be able to make it or not? Do you think you can make any project work? What makes the determining factor on whether or not you guys can help an organization a startup like Sebastian's? working with, get the job done. The hardest part of Health Care AI isn't the intelligence aspect. It's surviving the contact with reality. The projects that make to the production are designed around the real healthcare systems. So by real healthcare systems, I just don't mean the workflows which clinicians go through, but also managing the data. We all know one way or the other, we can get the data, but are the systems being built around latency of the data or missing pieces of the data? Because you know, all the systems work on different schedules. Some might have 30 minutes delay, some might even have overnight delays. So are the systems engineered well enough that they can handle those delays? And also in some cases, if they don't get the data, will they be able to manage their workflows and the outcomes? So that's where the, the S aspect, the signal aspect comes in. Do we have NF data? Do we have enough availability of data? Do we know the cycle in which we'll be able to find the data? So those are important things and co-founders. They need to understand that, okay, if I have to build this system which runs on data, if we get the data, we all know AI can run. But all the back-end aspect we need to happen to get the data and also account for lack of data is very important. And then once we have built a solid rail of data pipeline, then we do also figure out how it fits in with the workflow. How much changes are overburdened and overworld, clinicians have to change in their day-to-day life to get insights out of this application. And if it's too many changes, we know it's not going to be very successful or very well adopted. Other aspects which we have seen where product struggle are companies not realizing the budget data to keep in mind after products have gone live. You know, first few iterations of going live, which needs to account for what we call as agent drift. You know, given how fragmented and dynamic healthcare systems are, ecosystem is changing every day. ICD course change, regulations change, formularies change, you know, the lab standards change, which means there needs to be an ongoing and continuous monitoring of the performance of the CI systems and tweaks are needed. So we need to manage that agent drift as proactively as possible, which means not only building systems around it, but accounting for the cost for it. Also, we all know everybody is very, everybody wants to implement AI, but we also need to make sure when we're implementing AI, it's not adding more homework by bringing human in loop, you know, we all know he bringing human in loop is important, but we need to be very considerate about the fact how much and when. So if the outcomes which we are getting are of high confidence and they are well-sourced, maybe we need to build workflows that they can just pass through without bringing human in life, human in the loop. In healthcare, we are focusing on what problems we can solve with the power of AI models or the data I have, but the song framework is, I consider as analogy to the non-functional aspect that for your product to be successful, make sure you're understanding your data, not just what data you're getting, but the format you're getting in, what kind of normalization you need, what kind of changes it can have because of regulations or plans changing or formularies changing or ICD course changing, and also make sure you know how the workflows which are using it, what would be their needs for governance or reporting or how it will help the clinicians who are using it. So they are not challenged by the outcomes, they feel supported by the outcomes. Sebastian, could you just talk about how this is enabled the foundation of Charity to just be solid? I mean, I'm sure it's a co-founder to your point earlier. Like this integration piece is not the strength of the sort of core capability of Charter. And so having this sort of ally to Char's been in this for 15 years, like you said earlier, they know HL7 fire, they've got those guard rails for you to sort of just stabilize your foundation. Like can you talk about what that's done for you as a co-founder and enabling you to take the vision of Charity out into the industry? What the kind of partnership between Charter Health and Pegasus One really enables is it allows us to, I think from my perspective, really achieve our vision more thoroughly. I think there's quite a lot of different people out there that work on a variety of different things and are really great experts at things. I don't think you always need to reinvent the wheel in every single aspect, right? Like for Charter Health, we want to be really, really great at delivering these insights and analytics across these health systems and hospitals to be able to really kind of influence the way that these hospitals work, right? And for that, there's so many other components that are kind of foundational pieces like understanding how this interoperability works, like understanding data governance, like being able to make sure that agents don't drift over time. And I think those are all things that you can leverage other organizations and the expertise that other people have and have developed over time to be able to help you operate to a full capacity. And I think that's something that Pegasus One has allowed us to do. - Too short, like as so now you've got this foundation laid for Charter and Charter is launched and it's live and you guys are working well together. What benefits do you anticipate moving forward as Charter maturers and maybe scales into a bigger thing? - I would say as Charter scales for good or bad off it, it will bring new problems, which are great because so first of all, as it scales in my opinion, the immediate benefit for the users would be the time compression. So the research questions that are currently taking several weeks, somebody has to do data wrangling, they are exporting information to Excel, cleaning it, reformating it, and then running the analysis that timing should collapse from days to hours or minutes. So it will eliminate the friction between what insights I can get from this data to let me further explore the insights which I got from Charter. So long term, I think the benefit is what I would call compounding intelligence. You know, every Charter, every query with Charter handles, every co-authored bills, every analysis it runs, it becomes institutional knowledge. And as the patterns in image, future data preparations would be fast. So like the example Sebastian was mentioning, once they have done 100 subsistence analysis, the 100 one will be much faster and much more efficient. And that benefit will go to the researchers and the users of the application. And then will, I will say, we'll have a workflows multiplier effect. So since it's becoming more and more better and faster, so you need now less people doing low hanging work of data wrangling, data cleaning and massaging and being able to get more insights out of the data using Charter. Lastly, but I believe it's the most important thing which it would matter is the bottom line of the health systems, you know. So there's a compliance and quality angle to it. If Charter can help identify quality metrics proactively and surface cohorts for regulatory reporting automatically, that's just not efficiency. That's revenue production and risk reduction for our health systems. And I believe that's one of the most important thing, you know, health systems can extract out of Charter in addition to the insights. - So Charter's got a big vision for you Sebastian. What about you? What are you most excited about as you move forward? - I really like the Charter vision. I think it aligns kind of quite well with what we were seeing and kind of what our long term vision is for Charter Health and Charter as well. I think another thing that I wanted to mention was, there's kind of a term that gets thrown around, which is network effects, which is basically saying like, the more people you have on a system, the better, the system works for a lot of other people. And maybe you might see kind of a system like Charter as kind of individual people are using it and kind of individual hospitals are using it, but really kind of going back to Tishar's point, the way that we have it set up is it's able to learn and iterate and improve from those kind of individual processes and individual analyses that are run across these different institutions, right? And the more people you have running these analyses, the more kind of understanding you have from these different kind of ontologies that we have added in, these different. expertise that we have brought in from these different hospitals. Like, as we know, different hospitals have different ways that they like to do their own analytics. It's not kind of as much as we like to say, like, sepsis is the same across all hospitals. Every different hospital has different ways that they do their analytics, their understanding, their insights. All right. And the more exposure we have as an organization to the way that those are run, the better we can really kind of refine the analysis that occurs, right? So I think to go back to to Charles Point, the more the more occurrences that are run, the better the system operates and kind of the faster it's able to do that. The more accurate it is as well, what at the end of the day, Charter's goal is to really improve the quality of care across these institutions and these hospitals. And if we can really deliver those insights faster, then we're able to allow hospitals to implement those changes across their across the hospital more effectively, right? And I think one thing that I like to say sometimes is when you're, when these hospitals are doing these analyses, oftentimes they're doing the analyses on data that's multiple months old, right? They'll start a higher, some consultants in and those consultants will look at data from a few months ago. So let's say as an example, from November of 2025, and like right now it's, it's February of 2026. Within those four months, a lot of the people that are those cases of sepsis have either potentially left the hospital or are sometimes have died, right? And those changes that you're implementing are lagging kind of on that data that you have. And I think to be able to enable a hospital to say we currently have these 100 patients that are potentially either facing sepsis or we see that if we change these factors, we can potentially reduce their risk of mortality by 50%. Those end up being live-safed, right? And when you're looking at that data retrospectively, those people might not be around anymore, right? And if you're able to do that over the patients that were here in the last two days, then you can really make an influence on unsaving those people's lives. And I think at the end of the day, that's really kind of the goal for a lot of us. I think that's for why a lot of us go into healthcare, that's why a lot of us kind of are really working on the problems that we're working on as much as they are kind of interesting and great technical problems. I think at the end of the day, a lot of us kind of care about delivering better quality of care for patients, right? And I think the goal of Charter is as kind of Charter grows and as kind of Charter Health grows and continues to have our partnership with Pegasus 1 is really at the end of the day working towards that. I think the future of healthcare really depends on founders like you innovating Sebastian. So thank you for what you're doing. I can't wait to see what Charter Health does with the support, of course, of Pegasus 1. And I'll just ask to Charter, before we part, do you have any final thoughts for our audience? I would say every founder, and I'm talking even when they're trying to find a service where they need to be prepared enough to not only just focus on what they're building, but everything around it is more important because unfortunately, I've seen several clients reach out to us who have built great products but flimsy foundations. So when they, when any founder is finding trying to find a service organization to build their product, make sure they understand all the important aspects of a strong product which can make them successful. And it will be different industry by industry to some extent. There are some overlaps, but those foundational pieces are very important. Sounds advice. Well, thank you guys both so much for your time today. Always a great conversation. I look forward to seeing your presentation at the Vive event in the AI zone. And for more information on Pegasus 1, you all can go to Pegasus1.health for Charter Health. You can go to Charter Health at CHRTRHELTH.com. We'll put both of those links in the show notes to make it easy for you. And I hope that you all have a fantastic day.