Ram Yalamanchili (00:02.572)

Hey George, how are you?

George Magrath (00:03.905)

Hey, doing great. Good to see you.

Ram Yalamanchili (00:05.708)

Yeah, you too. Thanks for making time. And maybe let's start with a quick introduction about yourself and then we can start.

George Magrath (00:18.094)

Sure, so I am a physician, an ophthalmologist, and I have worked in industry for the past 15 or so years in different roles. I've worked as an equity analyst in New York. I've also worked in drug development and as a service provider. And in my last job, we were, there's a company called Lexitos, it was a small company that ran ophthalmic clinical trials. And we went on, we're running.

40, 50 studies. And I left there a couple, well, a year and a half ago and joined what is now Opus Genetics, which is a company that's dedicated to developing gene therapies for blinding conditions of childhood, right? So childhood blindness. So we have two, two assets in the clinic. One is for a condition called LC5.

The other one is just now about to start with clinical studies, which is best one. And so we're super excited to have a portfolio of gene therapies that come from the University of Pennsylvania.

Ram Yalamanchili (01:25.902)

Thank you. And what I find fascinating about your journey is the multiple roles and hats you've won throughout your career. I feel very related to that kind of a goal because I've also come from a very similar sort of multi-role perspective into the clinical trial industry. As you know, I've been a tech entrepreneur for the first half of my career.

coming from a computer science background and trying to break into healthcare about nine years ago is challenging. It's not the easiest market or the space to get into. I think like yourself, it requires a certain amount of expertise and training, which if you're coming from a pure compute background, it's not quite clear. But fortunately, what I've found is a opportunity to work with a

George Magrath (02:06.284)

you

Ram Yalamanchili (02:21.166)

few physician scientists like yourself as co-founders, I built a company called Lexand Bio. And what we've done at Lexand is develop a few molecular diagnostic assets in oncology. So the process of building a company like that, going from a small biotech company with an idea, fundraising, working through the product development cycle, your protocol development, your study development, recruiting, working with many

parts of the ecosystem like your CRO sites, various other vendors. So we did a, I would say, a full spectrum of all that work and I've certainly gotten to learn quite a bit about this process in the journey. And ultimately we had an outcome where we were able to exit the business to Roche in 2019. And I began my journey in terms of thinking about what do I do next? And one of the things which I was...

pretty fascinated about and quite sure about also at that time was the potential for AI to be a much bigger part of our thinking than it has been, say, in 2021 or 2020. And I also was looking around opportunities and I said, well, what is the area where there's the maximum impact because there's impact from automation and bringing AI into this mix?

And I think clinical trials, you would agree from your past as a CRO leader, is entirely a and human-intelligence driven market. And I think there's plenty of opportunity for us to really expand the bandwidth which this market brings. I think we should be able to do more with less, but at the same time significantly improve the quality, reliability, and consistency.

So yeah, so that's kind of where we are and what we're doing right now.

George Magrath (04:20.284)

Yeah, it's amazing. It's so complimentary what we're doing. My first experience with AI in clinical trials was really around imaging, right? We were using computers to identify patterns in images that might predict which patients would be the best responders to certain new medications. And that's super cool stuff, right? It's a very practical outcome. But what you guys are doing is...

incredible from an efficiency standpoint and from a quality standpoint because so much of clinical trials is sort of repetitive type work, right? Or work that can absolutely be handled by the computer. And so I know that our current trial working with you guys, that's exactly what we're doing and has led to faster startup, lower.

queries about every metric has been improved by using your system. And so I'm absolutely very grateful for your journey and getting into what you did because it's made a direct and immediate impact on our LC-5 trials.

Ram Yalamanchili (05:36.014)

No, it's really great to work with you and your partners or the sites you bring in. And yeah, we're very excited. feel what we're doing right now is really just the first innings or the second innings. There's so much to do, so much more to build on. And as we see, the intelligence is starting to really develop into a place where I think we can certainly see where this trend is going over the next few years. So let me ask you something,

Tell me more about your particular trial, maybe the opportunity, the design, how you're approaching it.

George Magrath (06:09.455)

Yeah. Yeah. So this is super interesting. This is really the core of the value proposition of AI or computer learning in our program, which is that we have a lot of programs. We have seven of them in our pipeline. But the trials are all very similar. It's just looking at different genes. in the world of inherited childhood blindness,

there's about 280 different genes that can cause this. And what we're doing at OPUS is we're targeting them one by one in a highly efficient conveyor belt sort of string of pearls type approach. And so for us, it's all about, it's almost like manufacturing, right? It's all about consistency and efficiency with these things. And so LCA5 is just the first and you can see our pipeline has seven. And so,

What I'm really hopeful for is that we can have significant cost savings, significant time savings by implementing with your system on the first trial and then just having the computer help us replicate it each time we do the same thing with a different gene. And so that's the real core value proposition of why we're particularly interested in this. And then to get to your question about exactly for the LCA-5 trial, really,

The interesting thing about this one with AI was that there are lot of different assessments in this one. There's a lot of different.

George Magrath (07:55.646)

second. There's a lot of potential variables, potential different data points that we need to capture. And the AI system was able to come in and absolutely give us a great first step on that.

George Magrath (08:24.646)

Yeah, a phone call came in and I silenced it and it gave me a 5241 thing again. Yeah, I'm sorry. Yeah. But so maybe I'll pick back up at.

Ram Yalamanchili (08:36.302)

Did we have to go?

I'm hearing an echo, George. Is that something anyone else is hearing?

George Magrath (08:43.828)

So I hear an echo too.

Ram Yalamanchili (08:50.03)

Can you mute yourself, Let me see if that's the reason.

George Magrath (08:55.33)

Do you still hear it? I don't hear it.

Ram Yalamanchili (08:55.485)

about that.

OK, it was up to. OK, cool. So I guess we can go back.

George Magrath (09:00.349)

Okay, so I'll pick back up where I was Abdul and maybe we can add it. So in the LCA5 trial and specifically, we're using AI and we're using particularly your platform for a number of reasons. And the first reason really is around the amount of data sets that we have, like the different data that we need to capture from all the endpoints and the ability of the computer to really be able to target in on that in a way that I haven't

really seen clinical trial databases ever do in the past, at least with the ease and the efficiency that you guys were able to do startup and able to execute. so that was really huge. The integration with the site is really great too. I've used that, as you know, on a different project in the past before OPUS, where we utilized that on a study that was more high volume, high throughput.

and the ability to integrate into the system, identify patients, give workflows to different roles within the practice made all the difference in the world. And as you know from that practice and that particular phase three study, I think they enrolled 100 patients in that trial. So was a very high throughput effort with your technology that I can tell you was

was made quite honestly was made possible by having something like this. It would have been extremely difficult. We would have needed multiple additional personnel to execute in a traditional fashion in that trial.

Ram Yalamanchili (10:43.266)

Yeah, and that's really the crux of it, right? I think when I look at the burden and the way we currently design trials as an industry, it's sort of lost in translation where I think we all design for what's best from our lens, but there's so many stakeholders in this whole ecosystem that it's very hard to sort of optimize or, you and I think it's very easy probably to like suboptimize for yourself.

George Magrath (11:01.434)

Mm-hmm.

Ram Yalamanchili (11:12.812)

I really think that's what's happening across the board. sites have massive burden on the execution side of these trials. Similarly, CROs or sponsors also have equally high amount of burden, right? Because your sites are your bedrock. And if sites are burdened, that burden does translate back into what I call upstream. So it could be monitoring, which is burdened because of the site's lack of streamlined work.

workflows. So I think one of the things I want to touch upon is because you've seen us and you've we've worked together both on a site side of things as well as the sponsor side of things where with Opus on the site side, I think I'd like to understand, know, as a sponsor, what are some of the challenges you've seen in the past and also as a CRO because you've you've you've read Lexitas and I'm sure you have a lot of perspective on

George Magrath (12:08.984)

Yeah.

Ram Yalamanchili (12:12.586)

running many, trials and working with hundreds of sites. So what are some challenges you've seen and are not addressed?

George Magrath (12:19.79)

Well, what you said, before I get into that, what you said first was incredibly insightful, right? So when we develop for a trial, we develop one system. But yet, there's so many different stakeholders that have so much different needs, and the ability to tailor to each of those makes everybody's workflow more efficient. And I think that's exactly the answer to your question, is that what I've seen from being on the sponsor side, from the CRO side,

And from the site side and from running trials on your platform is that it absolutely is, that is the answer, right? The ability to customize for different workflows is huge because as a sponsor, what do I care about? I care about enrollment rates. I care about queries. I care about getting the data and ensuring the quality of the data, things like that, right?

audit preparedness, all the quality metrics. From a site side, when I'm trying to enroll patients, know, as a PI, I care about, am I making sure that I'm doing all the right things in the right order exactly like I should? From a coordinator standpoint, she's trying to get in touch with the patients, figure out which patients on the schedule are the right patients to roll the trial, when do we do follow up, all the logistics of that.

Do we have the most recent consent forms? Are we doing everything like we're supposed to? It's incredibly executional. And then from a CRO side, you're basically in the middle trying to manage both of those. And so it really is an interesting concept to have one system that you program that then will pull out and create amongst itself.

different workflows for different users, different roles. So that's a very cool thing.

Ram Yalamanchili (14:22.318)

Yeah, and stepping back, I think when I was in my sponsor role at my previous firm, Lexand, what was fascinating is you ultimately were beholden to the performance of your site, or sites, I would say. And everything we were doing, for example, unlike my...

George Magrath (14:41.899)

100%. Yeah.

Ram Yalamanchili (14:48.224)

experience prior to Lexant, which was mostly tech entrepreneurship. You have certain metrics which are usually defined by product launch, some amount of early revenue, some ARR, some few logos of customers. And that's really how you go about and raise your capital and continue towards the journey of building the company. And my experience building Lexant,

has shown me that, you you really are beholden to your investors interest in the type of data you're able to present and also the wins of the market. You know, are you in a space where there is funding available or not? it looks, at least we went through these cycles where there were periods when you could raise and there were other where that whatever you're working on is not where the market is in terms of being able to fund you, right? So that was one interesting thought I came across and

The other thing on the site side is, obviously after starting Tilda, I took this approach of not trying to build a platform or any type of technology without fully understanding the whole picture, both sponsors, sites, as well as CRO. I have worked in a sponsor and a CRO relationship prior to this, prior to building Tilda.

But I would say the first half of our journey or the first third of our journey with Tilda, about a year and a half has been really working at the site level. We purchased a site, we went in and we really got to understand the challenges of running a site, what it's like to be a coordinator, what it's like to be a owner of a site.

George Magrath (16:24.615)

Mm-hmm.

Ram Yalamanchili (16:29.998)

And really getting at the nitty-gritty details of how do you manage 20 studies in parallel at a site, or all the issues you face. And in this process, I also got a chance to really get into the details of coordinating and working with the site staff, working with patients, talking to many sponsors, CROs, who would come to our site and talk about putting a study, potentially a feasibility, through our platform or through our site.

George Magrath (16:35.987)

You

Ram Yalamanchili (16:57.876)

And what's fascinating is so much of this is essentially up to the coordinator. I feel the coordinator is the most underappreciated role in this whole picture, right? The entire clinical trial picture. understand somehow that was just not part of the calculus when I was at Lexan. This was not a discussion we've had at our board meetings or our regular syncs within our team.

George Magrath (17:04.753)

Mm-hmm.

Mm-hmm.

Ram Yalamanchili (17:27.086)

where we said, you know, we have 23 sites or something, and how do we make sure all of our coordinators are happy? They're actually motivated, they feel empowered, they're going to do the work so that we can get the data. you know, I think having taken this other side of it, being a coordinator for about a year and a half at Tilda, I really started to appreciate a lot about the challenges, right? You know, I would say...

We had roles, multiple roles in a single job, I would say. You had everything from managing feasibility, budget negotiations, contracts, then ultimately doing your trainings, startup at the site, recruiting patients, calling all the patients, working through the backlog, bringing them in for screening, performing the screening, collecting the data, putting it into the EDC, doing all the quality and query management.

George Magrath (18:02.171)

Mm-hmm.

Mm-hmm.

Ram Yalamanchili (18:26.478)

Of course, there's regulatory, like you mentioned, the IRB communication and interaction for any new forms, new information, just coming in that way. And then finance, you know, ultimately getting paid for all the work you're doing, right? And I will say, going back and looking at our own processes and trying to do this with no AI or automation or what we call AI teammates, I just do not think it's...

George Magrath (18:38.17)

Mm-hmm.

Ram Yalamanchili (18:55.79)

It's a place where, to me, that was just unfathomable, at least where we are today. Because you can bring so much efficiency when you sort of AI into some of these day-to-day workflows. And you really collaborate with an AI. And some of these workflows are starting to become very, very intelligent just from where the feed was going. So yeah, no, I completely get where you're coming

George Magrath (19:23.848)

Yeah, know, the interesting thing is that you're exactly right. The site coordinators and the staff at the site are the ones who make the studies go. And they're the ones who produce the data that a lot of times your study falls upon, right? And then I've seen that over and over in studies, whether it's imaging data or ancillary things like that, that the technicians are doing.

and that the doctors were viewing and 100%. At sites I've been around and at the site I used to work at, you'd have a number of protocols on your desk and the coordinators were the ones who decided which one was at the top. And it was decided based on their ability to execute it and how comfortable they felt with it. And so it really is...

very critical to execution. Like it goes down, it goes down to the site coordinator and, and it's, it's, it's the, probably is the most critical position in a study, right, right there with the PI, you know, couldn't agree more. Yeah.

Ram Yalamanchili (20:36.46)

Yeah. And speaking of your own site experience as a PI and working on that phase three study, we worked on with your team, I think you had one coordinator managing what about a hundred patients at that point. And you are the top unroller in that study, right? To me, that was, that's really the power, right? How do we build that kind of bandwidth into this model?

George Magrath (20:56.887)

Yeah.

Ram Yalamanchili (21:02.754)

where you can enable sites to perform so much work at a very high quality and ultimately be able to really push the frontier of innovation. The current model is like burning down in many ways. I don't see how we can sustain this going forward unless you do something different.

George Magrath (21:21.004)

Yeah, I haven't even thought about that, but you're exactly right. We enrolled 100 patients across four surgeons with one coordinator who did all the work through your system. And it was the top enroller for that study. it's by far, like normally you would staff something like that with probably three or four people. So it really did.

increase the efficiency of our site coordinator by 3 or 4x. And that individual executed well, really enjoyed the trial actually, even despite that volume wasn't burned out, wasn't flustered, she enjoyed the trial. I mean, she'd have done another 100 if the trial had allowed it. So it really is powerful. It's very cool. Yeah.

Ram Yalamanchili (22:14.764)

Yeah, no, and I remember, you know, just the experience of working through that process and having these reactions like, wow, like, you know, I collected all my source data and from there on anything related to working with external systems, working with RV or monitoring, querying, a lot of that was happening pretty much through the AI teammate, which we were able to provide. And I think that's the power, right? And I don't see a world where

we all wouldn't have our own AI enabled teammates. I for what you do, what I do, and what certainly a site and a CRO and a sponsor would do. And I think that's really the vision where we're going, Batula. I think the opportunity to build right now is not just the AI, but the interface through which you can work with an AI. So what is the collaborative layer which you can build so that we can work with an AI, right?

I think traditionally, if you look at just, you know, hiring teams and working with within our own teams, as as humans, we have the capability to interact. We have emotions, we have personalities. We can work on different modalities. You know, we can share different applications on which we can bring context very quickly. These sort of things are not really there yet. When, let's say you use chat GPD or the chat interface is a very limited sort of an interface for performing real work, right?

or at least work which is related to the type of work we do in the clinical space. But I think our goal and what we've built is really just that, building a platform through which you can work with multiple AI-based teammates. You can collaborate with them, you can train them, you can teach them, you can monitor them. And I think I see a future where this sort of thing becomes normal across the board, right? I think it should be the way we expand.

the productivity we all have. And I think ultimately I want to touch back on something you said about the pipeline and the way you're thinking about Opus and the opportunities of running multiple different trials in a quick succession. And I think when we first talked about it, I was explaining to you the main motivation, at least for me, why we're doing this at Tilda is I really do think the current bandwidth

Ram Yalamanchili (24:38.242)

of how many trials we can run, how many patients we can recruit, how big the space for biotech innovation can be is really curtailed because we have not had any meaningful expansion in the amount of infrastructure in clinical trials over last 10 years. In fact, I would say, I would argue that the amount of infrastructure shrunk because we've lost quite a few physicians who would normally have done research.

George Magrath (24:53.981)

Hmm.

Ram Yalamanchili (25:03.502)

And there's data out there which shows that it's been going down or at least we have lost some during the COVID period and we haven't recovered back. And that is a problem for the industry, right? I think When we think about how do we push the frontier, I mean, especially now it's exciting because we have really strong models on the biology side, on protein folding side, like let's say alpha fold and models like that. So what to me that showcases is you are expanding the number of targets you can potentially

come out with at the very early stage of the development cycle. And that's great and wonderful, but once you get into the clinical model and into your human studies, where is the bandwidth? We currently do not have that bandwidth, and we need to do something to expand this 10-fold, maybe 100-fold over the next coming decade. And I think you're going at something which is very similar in rare disease, and yeah, we'd love to touch more on this, right?

George Magrath (25:59.777)

It's, this is the core of our business proposition, right? So we're developing treatments for rare diseases, right? For diseases that affect a thousand kids, you know, in the United States, like small stuff, but so meaningful because these thousand kids are going blind. And we know the technology works, right? mean, Luxterna was the first one approved by Spark and it worked. so, you know, running the clinical trials is

is so important to be efficient, right? If you're in a big indication where if you're after diabetes, obesity, something like that, then the numbers make sense in big pharma to go after those indications and run the trials with a lot of money and a lot of spend and a lot of time. But when we're going after these targeted diseases, in order to make it

In order to make it viable and to make the company real we have to be efficient and that's exactly why I'm so thrilled that you guys are doing what you're doing because otherwise these kids wouldn't get treated, right? I mean it just wouldn't be viable and so the bigger picture for the country or for the world is that the ability to use AI or the ability to use computer learning

in clinical trials to improve efficiency allows us to treat diseases we otherwise would not be able to treat.

Ram Yalamanchili (27:36.526)

What's stopping you from doing this without, let's say, AI or any of these latest technologies? For example, when you're in your CRO role, I'm sure you've dealt with customers like yourself, a pipeline of small, rare disease indications, or even studies which are not technically rare. But why are we not able to basically enable this opportunity?

George Magrath (27:51.81)

Yes.

George Magrath (27:55.213)

Yes.

George Magrath (27:59.213)

Yes.

I've watched a number of companies from the periphery when I was a service provider, you know, just not be able to get drugs developed because of the cost and timelines involved. Because it's just like any other business, right? And that's the thing is to impact these patients, you have to create a viable business or else it won't get across the finish line. So it's...

It's an interesting concept that not any other industry really has to worry about, which is that you're trying to do good for society by treating these patients. But you have to build a real business that's viable to be able to do that. And I've seen it so many times where it's great science. It really looks good. But the pathway

just isn't viable from a commercial standpoint and the project dies, you know, because it takes too long, it's too expensive, it's too hard to find the patients, you know, that kind of stuff. And that's fundamentally what I'm trying to solve from my standpoint is how do I run these trials as lean, efficiently, with as high quality as possible, right? And we're...

Absolutely 100 % using every tool we can and a big part of that is AI and through you guys, honestly. That's how I look at it. If I'm developing these seven projects and I'm developing them at price points or at...

George Magrath (29:49.91)

timelines that are used by bigger indications, then they may or may not make sense. So it's all about efficiency, getting these things to patients, proving they're safe or efficacious, and then trying to get them approved.

Ram Yalamanchili (30:08.174)

And I think I found something interesting in what you said, right? It's not just about how much your overall budgeting is, it's also the timeline. And I think one other thing which I want to point out is it's also about how you can paralyze these multiple approaches at the same time. And

George Magrath (30:25.397)

Mm-hmm, mm-hmm.

Ram Yalamanchili (30:26.934)

I think you could argue that yes, you might have the resources, but how do you paralyze seven trials at the same time and sort of go at it at the same pace as, you know, running like two or three trials, right? And do you find that as as another sort of lever you need to think about?

George Magrath (30:42.823)

100%. Like most companies my size, you know, can concentrate on like one asset. And we are just concentrating on one right now. So, but as this thing unfolds and if it does what we hope it does, the goal is to unlock the ability of a small team. You know, we have a small team of people, a small team to be able to execute.

in parallel the number of trials, right? And that's what will drive value for our shareholders. That's what will get treatments to patients faster. That's what will move science forward for this. And so that's absolutely the goal. the answer, you know, for a small company, the answer can't be, well, if I need to run seven trials, I'm going to hire seven times the number of people. That's just, that just isn't a realistic business proposition. It needs to be, how do we leverage what we have?

from technology to increase the efficiency of the team that we have. So that's what we're trying to unlock. It's happening. I mean, in real time, like we're seeing that. Yeah, it's very cool.

Ram Yalamanchili (31:50.816)

It's also not guaranteed that if you do expand your resource by seven times that you actually have seven times the bandwidth, right? I that you probably have some optimal scale at that point.

George Magrath (31:58.18)

No, and as you know, you expand like that, you start to have to add in layers for management oversight, things like that, quality that come with having that many more people. Absolutely, yeah.

Ram Yalamanchili (32:16.974)

Yeah. And just touching on the aspect of quality, I have seen how traditional monitoring works, the traditional service industry works around this. And as you know, it's entirely driven by a person going to the site or maybe going to the systems and

looking at all the documentation, all the data which is coming in, and really doing like forensic level monitoring on top of this. And I'm just curious from your perspective, how do you look at the role of AI going forward, right? Because I'm sure certain parts of it cannot be done through AI, and some can be only done through an actual monitor who's on site or with the site.

But how are you thinking about that going forward in terms of the way I can take this?

George Magrath (33:08.293)

Yeah, well, so the advantage of, I think one of the big advantages of using computers in this is it is really pattern recognition early in real time because what we're trying to do, it's a quality first of all is the number one thing for us, right? If we're not running high quality trials and the rest of it doesn't matter. So quality is always number one for us and quality to us.

will likely always involve humans to some degree. But what saves me time, what saves on quality, what's best for patients really, honestly, at the end of the day, is if you can recognize patterns in poor quality very quickly and be able to mitigate them very quickly.

A computer does that in real time with the data. The traditional way to do it with monitors is very hard to do it in real time. And one of the first applications that I used AI for that was for a different company that was more image-based was looking at the quality of images in real time. The patient is sitting with the coordinator in front of an imaging machine. The image gets taken. And before that patient even gets up out of the seat,

the quality is reported back to the coordinator. So if the retake needs to happen, it happens at the point of care, right? So it's not a human looking at it that afternoon and then they have to call the patient to come back and retake the images or anything like that. It's happening in real time. And doing that throughout the data set is critical because the worst possible thing you can have happen, maybe not the worst possible thing, but one of the bad things that can happen.

is if you get systemic patterns of poor quality in a data set that isn't corrected quickly. And so that's a huge, huge thing.

Ram Yalamanchili (35:17.196)

Yeah, and what you're pointing out is actually very interesting because we see this often with sites we work with as well because the monitor is doing quality management on behalf of the sponsor. But I think the AI would also need to do the same on their behalf for themselves because you are ultimately responsible from the quality of work you're performing from an FDA perspective. And we have an inspection readiness product where we have an AI teammate which is like,

quality teammate and is essentially checking all the work being performed at the site and essentially in real time reporting where you are with your inspection readiness. And if there's any like, you know, areas which need to be buffered up or there are areas where we need to go back and like you said, maybe an informed consent was not the right version and you had to reconsider that patient or it could be a missed report somewhere, which is not part of your binder, not part of your data set. And so...

What I've found fascinating is even in the most incredibly efficient sites and places where they have a really tight quality program, you still cannot get to the accuracies which we're getting through the current AI-based implementations. And I think it's just not on the quality side. I've found that even on the finance side, for example. And it's fascinating because one of the

One of the articles I read recently said something like 95 % of the sites do not bill for all the work they've performed. And it's reasonable, right? Because in an industry like healthcare, I think we've evolved very strong systems over the last few decades to be able to solve for that exact same problem, which is billing. And you have this EMR infrastructure, which...

George Magrath (36:56.768)

Yeah.

Ram Yalamanchili (37:11.352)

I think some would argue, physicians like yourself might argue, the EMRs are actually more billing systems rather than healthcare, patient record systems. And we don't have anything like that in clinical research. These stakes are high and it's very easy to miss a certain set of invoices or certain set of visits you've performed, unscheduled visits, certain drugs you might have used which need to be passed through to the sponsor. So it's just a...

know, everywhere we look in clinical research, I find the burden to be on somebody to make sure like everything's going well. And it's a, I don't think it's a sustainable model. hence we see this whole concept of burden, right? Whether it be sponsor CRO side or site side. And then just sort of the whole process down for all of us who are ultimately trying to get more, more shots on the globe.

George Magrath (38:03.764)

Yeah, your point on the finances is so pointed. It's so true. The clinical side of things like the clinical care, the EMRs do such a great job of capturing, I mean, that's one of the biggest advantages of it, honestly, is that they capture the billing and they capture how you're supposed to appropriately bill for the work that has been done. And on the clinical trial side, that's...

completely ignored by the software to the extent of my knowledge. And that burden is put on people. And I think that is something that's probably a very low hanging fruit for sites to be much more efficient.

Ram Yalamanchili (38:51.584)

And talking about the imaging-based AI you've used in the past from a quality perspective, I find it fascinating because today the problem, least from looking at it both from a biotech perspective as a biotech founder and now from a site perspective,

George Magrath (38:55.389)

Yeah. Yeah.

Ram Yalamanchili (39:10.286)

The place where you can give me a technology like that as a coordinator and for me to actually use it just does not exist right now. None of the systems I have used or I can, know, Sierra would possibly give me or a sponsor would give me has that capability where, you know, you can bring these external integrations very quickly and enable you to do things which would otherwise be done manually. And I think, feel like everybody benefits from this, right? So what's missing here? Have you looked at ways to bring, you know,

George Magrath (39:17.768)

Mm-hmm.

Ram Yalamanchili (39:39.5)

transfer to technologies like that into the clinic or into the place.

George Magrath (39:42.439)

Yeah, you know, it's a wide open field, right? I mean, we did it with a very homegrown solution. And to be quite honest with you, I spent most of my time on an imaging problem that's slightly different. You're talking about a very practical problem that needs to be solved. The problem that I've spent most of my time on though is...

in an era of personalized medicine, this is the exact tag that this goes under, how do you make sure that the treatment you're giving has the highest likelihood of affecting the patient positively? And so can we use the computer to assist the physicians, assist the investigators, assist the sponsors in selecting the patients whose biology is most amenable to whatever treatment you're evaluating? And we're actually, we actually just got it.

some of our work on that in diabetic eye disease accepted by the American Journal of Ophthalmology. So we're publishing on that very shortly. every patient is different and everybody's biology is a little bit different. And we have treatments that work for some people but don't work for others. And you really want to just not treat patients who are not going to respond and make sure that patients who will respond have access to whatever the treatment is.

That's the goal. I think that using imaging and using computer vision and computer learning for that is one of the, it's one of my passions, right? It's something I think will be so good for so many people if we can actually figure that out.

Ram Yalamanchili (41:23.628)

Yeah, and I think to your point, right, there are amazing technologies out there. There's people working on these problems which are clearly published and they have great, I would say, benchmarking looks good. But pulling that into a trial pretty easily and building a platform which will allow sponsors to do that or even developers to put it onto a platform so that you can be part of a trial.

George Magrath (41:44.441)

Yes.

Ram Yalamanchili (41:48.344)

I think that marketplace style platform where you can very easily like design your trial, pull in the right components, the right teams, collaborate with them. All that seems to be missing in this industry. Yeah, exactly, yeah.

George Magrath (41:50.426)

Yeah.

Yes.

George Magrath (41:58.501)

It's it's like the app store, you know, like like you go like you go in the app store and you're like we want to use this this and this. Yeah, that doesn't exist and it's hard. Like when I was on the CRO side, it was almost like you hit on you hit on something that's so real. There's so many different technologies out there, but actually implementing them in a clinical study gets extremely complex very fast for a myriad of reasons. And so having a platform.

where you can actually have like, you know, and you'll know this better than me, but like an API or some sort of plugin, you know, my very crude way of calling it an app store is sort of how I envision it. But that would be so meaningful for clinical trials. know, we do it in rare, in rare disease. We see it because we're using really specialized equipment that's targeted towards that particular pathology. And it may not be widely used because it's only applicable to a certain number of patients.

And so having easy interoperability is an unsolved problem right now.

Ram Yalamanchili (43:06.006)

Yeah, no, absolutely. And I think a couple of things which come to my mind when you're saying that is, I think ultimately you can do all of this if you had enough resources. I mean, that's well understood. But being able to go from a protocol to a first patient, first visit.

in a time of, a matter of, you know, hopefully a week or two, right? Something like that. In a very short timeframe where you, you've used majority of your resourcing to really like check and validate and make sure whatever your platform or AI has sort of set together. And then being able to run multiple studies in parallel with this architecture, right?

George Magrath (43:27.693)

Yeah? Yeah.

Ram Yalamanchili (43:45.974)

And then you have a series of workforce, which I call AI teammates, constantly checking, making sure everything's going well, making sure your quality is on right place, making your startup is on right place, and then collaborating. I think all of these seem to be the type of things which we should shoot for, and that's what we're going for. And

And ultimately, the way I look at it as companies like yourself will be able to deliver and effectively compete in a world where you're saying, like, I've got seven assets and I want to take it to the trials and I want to deliver value to my shareholders. yeah, I mean, you can go two at a time or you can go seven at a time with this. Hopefully, the of resources are, or even the resources actually feel at least on the capital side.

George Magrath (44:30.571)

Yeah.

Ram Yalamanchili (44:35.742)

I feel like people or companies which differentiate with their strategy and can actually effectively show that they're not distracted when they run these many programs probably can raise a lot more capital anyway. There's probably enough capital that's on a risk adjusted basis. It doesn't work out if you're distracted and you've got low productivity.

But if you can unlock that productivity and you can show that you as a platform or a team can manage multiple assets, and hence you need to raise more, I feel like that shift might happen very quickly, I would think, just in the PyTex case.

George Magrath (45:10.336)

I think it could, right? I mean, you prove that and then the capital allocation will be there, right? And it just goes to show you that the power of this is even, so far has not been unlocked, right? Because people haven't been able to really execute against it. And so,

the new technologies and stuff that's being built. I mean, how amazing would it be if every biotech out there could do more assets or do more science than they're currently doing? I mean, you would see such an amazing impact on our society if that happened on large scale.

Ram Yalamanchili (45:59.342)

And I believe that those days are not too far away, George. Even if you look at some of the offshoots of Alpha Fold, Google's products, I mean, that's what they're saying. They're saying we will have a drug for every target, almost, right? And that's great. think in a science perspective, maybe we can simulate certain parts of biology and figure out the right protein or the right

George Magrath (46:09.833)

Yep. Yep.

Ram Yalamanchili (46:29.014)

antibody design, but bringing it into the clinic, bringing it into the research pipeline, and then ultimately seeing that kind of maybe a thousand, 10,000 fold expansion bandwidth.

I'm looking forward to that day because I think we would need, certainly we would need many more people working in our industry when that happens. And we certainly will need equally number of AI teammates working alongside us. So I really do think it's an age of abundance, which can get unlocked if all of this plays out and we go to a situation where we're talking about. It'll be an interesting world.

George Magrath (47:02.846)

Mm-hmm. Mm-hmm.

George Magrath (47:11.303)

Right now the bottleneck is in clinical development, right? You've got Alpha Fold and many others that are identifying novel targets and identifying how to drug targets we thought were undruggable. But right now we haven't seen the same sort of efficiency gains on clinical development to match that. that's exactly what you guys are trying to unlock. I mean, it's super cool and it could be very, very impactful.

Ram Yalamanchili (47:41.066)

Yeah, no, I would say I've certainly seen different approaches to solving this particular problem, right? Clinical development certainly is a bandwidth constrained space. And I think you could argue maybe we don't need the number of patients we try to recruit. Maybe there are better ways to design a study or use synthetic data, synthetic patients, control arms. I'm sure you've seen all these sort of ways to reduce the amount of work we would need to do.

George Magrath (47:56.38)

Mm-hmm.

Ram Yalamanchili (48:10.198)

on the clinical development side. But where we are focused and where I'm passionate about is even if you were to do that, I still think there's a hundred times, thousand times for increment in the amount of work we would need to do in clinic, as in like with patients and with clinics and with physicians. And that simply is just not possible right now. Everything right now seems at the brim of whatever capacity we have and hence we need some new opportunities to really improve the bandwidth.

And if you kind of look at our way of bringing AI into this field, we think about it from a perspective of how can I improve the bandwidth of a certain persona? So if it's a site, it's a coordinator, it's a quality person, it's a regulatory person, it's a finance person. If it's a sponsor or a CRO, we have different personas. They generally have data managers, project managers, site activation. And so we've gone through that process of

George Magrath (49:03.717)

Mm-hmm.

Ram Yalamanchili (49:09.206)

identifying personas which are operationally very burdensome and bringing them in and training certain aspects of their workflow through an AI teammate. essentially saying, this is how you expand the bandwidth. Now you can do 10 times more work. And I'm very positive. think in general, this trend will continue to develop because there's a huge, I sort of say the world's mission today seems to be the pursuit of AI singularity.

Everywhere I look, all the resources and all the smartest people seem to be working on some sort of AI to deliver a very superior intelligence. So it's going to happen. And I think we all need to be prepared to take advantage of this new model, right? And go from there.

George Magrath (49:59.161)

100%, yeah. And we've loved working with you guys. I think we've got an amazing, it's just an amazing fit with our pipeline and our model, right, where we're developing an ultra rare disease to unlock efficiency with you guys without sacrificing any quality. And so it's fun to be at the tip of the spear, yeah.

Ram Yalamanchili (50:03.938)

Yeah, my question.

Ram Yalamanchili (50:28.17)

I'd love to see 70 assets in your pipeline,

George Magrath (50:29.206)

Yeah. Yeah, me too one day. Yeah, one day.

Ram Yalamanchili (50:35.47)

Yeah, I think you told me that there were, what was it, like 200 plus targets which are, the opportunity is large, right, from a rare disease perspective.

George Magrath (50:42.584)

There's enough of these genes that need to be developed to keep us busy for quite a while.

Ram Yalamanchili (50:51.282)

Yeah, it's so motivating. There's some fascinating amount of opportunities which get unlocked and the benefits and the impact is incredible.

George Magrath (50:59.222)

Yeah. You know, if there's ever any question about how important it is to be doing this kind of stuff, I would encourage you to watch the KOL event, the physician event that our company did in December where they described in detail the life-changing effects in the first three patients we treated in the LC-5 program. it's traumatic and it's motivating.

It's, you know, we got 280 to go.

Ram Yalamanchili (51:34.19)

I have a fun anecdote there. When I first heard about your company, I heard through our CRO partner who's working with you. I Googled it, and then I found this New York Times article. And I think the headline, or at least the abstracts, something like, somebody who had no vision all their life was able to see for the very first time.

George Magrath (51:58.186)

Yeah. Yeah.

Ram Yalamanchili (52:00.178)

And I was just thinking, this is like sci-fi stuff, right? This is unbelievable. And I read through the article and I was just completely blown away. was just like, how is this even possible? How can somebody who's been blind all their life? And I then started thinking about what are the implications then, right? I mean, what happens when you, for the very first time in your life, actually see it? And how do you process all that? must be...

George Magrath (52:12.288)

Yeah.

Yeah.

George Magrath (52:19.082)

Yeah. The first patient in our trial was treated a year and a half ago now, and he was 39 years old when he was treated. He had been blind since he was one, and he got formed vision for the first time in his life. He learned what a ceiling fan looked like, what a skateboard looked like. mean, was cool. Yeah. Yeah. Yeah.

Ram Yalamanchili (52:44.758)

Yeah, I mean, just hearing it is like such an emotional thing, right? So it's great. I love the work you're doing and we're so happy to be part of the journey you guys are on. No, this is frankly why we're doing it. This is motivating. We've to keep pushing and work on this. yeah, thanks, George.