AI in Clinical Research: 1 Platform for 30+ Fragmented Tools

[00:00:15] Ghazenfer Mansoor: Hello, I’m your host, Ghazenfer Mansoor. On this show I get to sit down with entrepreneurs, founders, and business leaders, talk about the bold decisions, pivotal moments, and innovative ideas that shape their journey. This episode is brought to you by Technology Rivers.

At Technology Rivers, we bring innovation through technology and AI to solve real-world industry problems. If you’d like to learn more about our work, head over to technologyrivers.com and tell us more about your project.

Today on Lessons from Leap, we are joined by Dr. Harsha Rajasimha, founder and CEO of Jeeva Clinical Trials, an AI-ready platform transforming how clinical trials are designed, managed, and delivered. Dr. Harsha, welcome to the show.

Why don’t you just introduce yourself and tell our audience about your journey, how you got to where you are, and then we’ll dig deeper into it.

[00:01:04] Harsha Rajasimha: Absolutely, Ghazenfer. Thanks for having me today. And like you mentioned, I come from a background in computer science and computational biology.

Much of my career I have spent at the intersection of computation, biology, medicine, and drug development. And after my professional experience as an NIH and FDA trained data scientist and precision medicine space, my personal experience of having one of my three kids born with a rare disease in 2012.

And so my personal and professional experience just converged in some unconventional ways that gave me a direction and purpose to apply everything that I had learned towards speeding up therapies, because there are more than 10,000 rare diseases.

Ninety-five percent of them do not have a single approved treatment option as of now, even today. And thirty-five years ago in nineteen eighty-three, there were hardly any drugs available for any of the rare diseases. Today we have eleven hundred approved orphan drugs.

And now with the advent of AI, genomic sequencing, gene therapies, and many other advanced technologies, we are able to make new drugs for many genetic and rare diseases, which would not be done even a few years ago.

So I am excited to be on a mission over the last seven years to reimagine the infrastructure supporting discovery and development of new therapies, particularly for rare diseases, but also for common diseases, so that a lot of humanity is at risk right now.

As many times when babies are born with a rare disease, the best thing doctors are able to say is, “Sorry, go home, and it’s bad luck,” and wait it out, right? And that’s not acceptable, especially given that much of the science and technology already exists to solve thousands of these diseases.

Already, the economics and the financials are not in the favor of the families affected by these diseases. But collectively, there are more than 300 million people affected by rare disease worldwide.

And so that would be the third largest country in the world if we put all the patients suffering from rare diseases in a country. And that’s not a small number. It’s almost like diabetes.

And so it’s essentially a mission for me personally as a social entrepreneur to do whatever best we can to catalyze, to ensure by twenty fifty we want to see all human diseases treatable with at least one treatment option. And I think it’s doable with the technologies available today.

[00:04:10] Ghazenfer Mansoor: Yeah, no, thank you. Thanks for sharing that. So can you talk more about what the Jeeva clinical trial does?

Obviously, the clinical trials are essential in the patient treatments. So, but the process is very fragmented. So what were the gaps that you noticed that led you to start Jeeva trials?

[00:04:32] Harsha Rajasimha: Yeah. So as I was consulting with many big pharma companies in the New Jersey area for several years, I noticed that the infrastructure supporting drug development and clinical trials was heavily fragmented. There were at least thirty different tools and IT systems, data systems and silos.

And then there is a lot of friction where for people who are involved in the clinical trials and drug development from wearing multiple different hats, in research, in clinical trial of between the biopharmaceutical company, a contract research organization, or a CRO, academic medical centers and clinical trial sites, and ultimately patients and families that participate in this.

All of them have to collaborate in multiple ways between clinical trial managers, data managers, medical monitors, clinical research associates, coordinators, investigators.

There are twenty or more roles that are involved that have to play their critical functions to make one drug successful in a highly regulated and compliant manner, without compromising patient safety, without compromising data security and privacy.

So it’s a tall order and a tall task, especially for an innovative new startup to come up and establish in this market is a huge endeavor. And what we saw is that there is no dearth of point solutions, right?

So a small tool that does a very specific function, which makes it necessary that that small tool gets integrated into a slightly larger platform, and then that becomes part of a much larger platform, etc.

And so it’s a very complex infrastructure that’s designed to remain inefficient, expensive, hard to maintain, hard to upgrade.

And so as we went through pandemic, after I founded Jeeva Clinical Trials and after we saw the launch of AI and Chat GPT and the large language models, we have gone through a couple pivots, and today we are in a much different world than we lived in a few years ago when we first started.

The vision and mission has remained the same.

What we are delivering today to many of our customers is a single unified software as a service hosted on the cloud, powered by AI agents, fully reimagined from the ground up in a way that helps them go from the start of the trial, which often begins with clinical trial protocol design, and take it all the way through database lock and ultimately to regulatory submission for marketing approvals by the FDA or similar agencies worldwide.

And in the case of rare disease, oncology, dermatology, we have demonstrated that the platform is validated, it works, and it’s delivering significant cost savings, time savings, and burden on humans.

Because the AI agents can do a lot of the busy work that humans can focus on decisions rather than looking for errors in the data, inconsistency, anomaly, and trend analysis, which AI is much better and accurate in doing this without getting bored or frustrated.

Whereas humans are likely to get frustrated, they may even be in a good mood one day and in a bad mood another day. But AI is consistent if designed well, of course, and we are seeing that it’s delivering what we designed it to deliver, and it’s exciting.

[00:08:42] Ghazenfer Mansoor: Yeah, is it primarily for the organizations running the clinical trials? Is it for the patient? Is it for both? Is it for caregivers? Or do you have the interface for all the different…

[00:08:56] Harsha Rajasimha: Yeah. Great question, Ghazenfer. So everybody has their own personalized interfaces so that all of them can be users of the system.

The customers ideally are the biotech, pharma, med device development companies, emerging small biotechs, or phase one, phase two clinical trials. That’s where we have had the biggest success till date. Now some of these biopharma companies have small teams.

They run it themselves by hiring some external CRO or contract research organizations that fill some of the internal gaps they may have.

And Jeeva itself, with our AI agents, with very minimal human staff, we can support the execution from designing and optimizing the protocol to database lock with biometric service, all in one turnkey solution.

Some of our customers tend to have their own CRO that may take over some of the staffing function and monitoring function. And sometimes the sponsor does it all themselves using the Jeeva software. So we have had a mix of both.

Essentially, it’s a software as a service, subscription-based model on a per clinical trial or an enterprise level subscription. And they only pay as they scale. If they are starting off with zero patients, as they enroll more and more patients, it’s a per patient per month model.

And that’s also a complex endeavor to provide pricing and cost estimates and budget for clinical trial execution.

Many of the CROs and our competitors tend to require a confidentiality agreement, exchange protocol, wait for two weeks, they analyze, and then they come up with a ten page proposal or a fifty page proposal.

With Jeeva, they go to our website, no confidentiality agreement needed. They can directly enter some of their general high level information, like number of patients, number of sites, number of countries, and they quickly get a price immediately on the website without having to share any confidential information.

Now, if they like what they see, of course we work with them to get a confidentiality and exchange protocol. And the first thing we do is… a lot of the mistakes in clinical trials are repeat mistakes. We see the same problems every single time.

The same protocol amendments, multiple times they have to amend the protocol, which is expensive. There are ways in which we can avoid and minimize the number of times a clinical trial protocol gets amended. The second is enrollment delays.

As most people in the clinical trial world know, patient recruitment is the biggest bottleneck in clinical trials.

And so if we can help speed up patient recruitment, they usually come and ask, “How can you help us speed up patient recruitment?” when there are problems upstream while designing the protocol itself. So we like working with sponsors when they are designing the protocol.

If they come up with a draft protocol, we have the AI technology called Clintelligence, or Clinical Intelligence, that analyzes about 600,000 clinical trials on clinicaltrials.gov based on real-time querying of the database to understand what aspects of the clinical trial protocol need to be optimized so that we avoid all the costly mistakes that have already been made by past clinical trials.

We can make some new mistakes. Why make the same mistakes again and again?

And even in this day and age of AI, we see that customers themselves tell us that, Harsha, we see the same repeat things going on at CROs or at biopharma companies, large companies that have run hundreds of trials in the past, they don’t feel like they are getting more efficient and much better.

And that’s frustrating to the whole industry, which has left the industry with significant costs. It costs about two and a half billion to make one drug, because nine out of ten drugs fail.

And it takes 10, 12 years to bring a therapy from first in human clinical trial to regulatory approval, that’s way too long. And so how can we speed up and reduce cost? It’s one, smarter design upfront.

Manage the complexity, identify all the common mistakes that have already been made, avoid them upfront. Using AI, we give them critical warnings and critical recommendations.

It’s up to the sponsors if they want to take it or leave it, the clinical researchers, the principal investigators, they can make their decision. But our AI gives shockingly good recommendations.

Our customers express that, wow, this is something someone with thirty years of experience running drug development and clinical trial programs would tell, maybe a couple of these, but this tool gives us everything in two minutes, all upfront.

[00:14:39] Ghazenfer Mansoor: No, no, this is good. Thank you. So I have two questions on this one. I’m gonna come back to this, how do you get the accurate result?

But before that, you mentioned earlier that with the arrival of Chat GPT, Claude, and all of that, you pivoted your product. So what is that pivot? Can you talk about it?

How did this new approach, new LLMs, cloud-based LLMs, change your strategy?

[00:15:08] Harsha Rajasimha: Absolutely. Yeah, I think it’s changing the whole world, and we are no different. And essentially what we had been focused on was the first problem, which is the fragmentation problem.

Too many point solutions. We built a unified infrastructure from the ground up. It was all built on Python, Django, React, modern technology stack in the last five, six years, and that was solving that one problem of fragmentation.

We were unifying the infrastructure. Now when AI and ChatGPT came about, and we started looking at Anthropic Claude and AWS, we were already on AWS Cloud technology stack wise. And the Bedrock framework was launched in twenty twenty five May for life sciences.

And so as we looked at how quickly it was getting matured, when it was first launched, it seemed like the technology was there, it was not fully mature. And even today it’s getting better and better, but it’s way more mature today than it was just like nine months ago.

So what we leveraged is one, real time analytics of just using APIs of clinical trials.gov to understand what were the common protocol inefficiencies of the past six hundred thousand protocols.

Correlation between specific elements in the protocol and specific inefficiencies like protocol amendments, enrollment delays, and several other types of inefficiencies.

So using that to bring that intelligence to that one trial that our customer is trying to optimize, what can we learn? And what are the most similar clinical trials to this one trial that we are looking at?

We don’t want to bring a trial that was run on a completely different disease area in a completely different geographic location on different patient populations with different endpoints, et cetera. We want to bring the most similar identical trials. What can we learn from them?

In similar geography, similar disease area, similar phase of trial, et cetera. And similar modality. If it’s a small molecule, biologic, a gene therapy or a cell therapy.

So using all of this in real time, it was not even possible before the AI LLMs came about. And so we essentially are delivering things that are humanly not possible. We are not simply looking to find areas where we can replace human activity with AI activity.

These are things that humans can’t even do.

Even the smartest among us cannot possibly analyze 600,000 protocols, identify the hundred most closely matching protocols in real time, and say, “Here are your critical mistakes in this protocol that you can fix.” That’s one part.

The second part is of course reducing the burden, right? So humans, even if they can do it, how many of us will enjoy going through the same thousand mistakes and correcting the same thousand mistakes every single day of our life? It’s boring.

And we would rather focus our intelligence on making decisions and choices and let AI do the routine, mundane, boring tasks. That’s exactly… But these boring tasks are what makes the data quality better, data integrity more reliable, and accurate, accuracy of the data.

And that’s basically what some of our AI agents do once the trial begins.

On an ongoing basis, as soon as we enter new data of a new patient on a new visit, it immediately takes that record and looks at all the past records of that same patient and looks for inconsistency, anomaly, errors, and contradictions.

In one place, we said this is a male, in another place, we are saying this is a female. In one place, we are saying this is age five. Another place you are saying age is 15.

Which one is it? For a human to even recognize that there was an inconsistency, because at today I’m only entering one data. How do I know what data was entered last visit, last month?

Unless I go and look at that data set. AI, it’s doing this in real time. As soon as I hit submit, it immediately flags a query, like, “I see that you just mentioned the age is fifteen. Last month you said your age was five. So which one is true?”

So I made a mistake. It’s five, it’s not fifteen. So I go and correct it right there, right?

And so those types of historical range, trend based longitudinal analysis in real time is time saving, cost saving, data quality, and even the whole clinical trial is at risk if we do not apply AI the right way.

So we are very excited to see what the potential of what AI is delivering in the last nine months, that our customers and clinical researchers are just very impressed.

[00:20:39] Ghazenfer Mansoor: And you raise a good point of the human in the loop, because as you are entering different sets of data, now it’s picking up those anomalies and giving it in front. So obviously there’s a big talk about Agentic, we talk about AI just making decisions versus, okay, the human in the loop.

And this is, in our business, which is also, we’re building AI in the healthcare space, we talk a lot about the human in the loop option because without the human reviewing the data before, whether it’s the medical data, you can’t just give a recommendation to the patient without doctor viewing it and making it, adjusting it.

Because there are, even irrespective of how good it is that, you still need to have a human validate, because that’s the trust part. It’s gonna take a while. So is that the same in your experience?

[00:21:31] Harsha Rajasimha: One hundred percent. Is such a highly regulated space with good clinical practice, good laboratory, good manufacturing, GXP environment as they say it, and it’s highly regulated by FDA in the US, European Medicine Agency in Europe and CDSCO and in India, for example, and many other countries have their own regulatory agencies.

And so when operating in a regulated environment, following data security standards, electronic data capture like the FDA twenty one CFR part eleven, HIPAA data, patient data privacy, patient data security, protected health information, and in a regulated environment means significant expectation on data security, privacy, allowing role-based access controls, and two-factor authentication, multi-factor.

So some of these are generic across various industries, and a lot of it is domain specific in clinical trials.

So having an understanding of the global standard of good clinical practice, what is called the ICH GCP or the International Code of Harmonization, good clinical practice, is critical in running clinical trials the ethical way.

And that means whether we use software, AI, digital technology, or do it in paper and pencil, which some parts of the world, even today, use the good old paper and pencil technology. It works.

It takes a long time to transcribe all of that into digital for doing further biostatistical analysis after all the clinical trials are completed. They need the data in computational formats to perform biostatistics and prepare the tables, lists, figures, and then prepare for regulatory submission.

FDA was accepting paper-based submissions until a few years ago. And literally an 18-wheeler truck load of paper would go into an FDA doc for just one drug, one regulatory submission. And not sometimes, and this is not even a joke.

This is real FDA staff validating that they have multiple trucks going just for one drug for a complex large global clinical trial…

[00:23:59] Ghazenfer Mansoor: Wow.

[00:24:00] Harsha Rajasimha: …coming from multiple different places, boxes, boxes, reams of paper. And so that’s just not practical. Now FDA itself has modernized and upgraded to electronic submissions.

So most of the regulatory submissions today are electronic, which means paper and pencil, while it works, is no longer preferred and acceptable, because anyway they have to invest later in converting that into digital, which is going to be more expensive and take up a lot of time.

So start right from day one, start on the right footing with the modern technology stack, which is ready, user-friendly, and globally scales, so that data is all collected in electronic format, so that the rest of the workflow is also Agentic AI driven in going from after the database is locked with the last patient’s data, all of the data can then be taken through standardization.

The global standards for CDISC, CDASH, SDTM, ADaM, doing the biostatistical analysis, tables, lists, figures, and then CSR or the regulatory submission report, the way FDA would expect, that we prepare the documents and data in a way that can be submitted after a human review, of course.

And so to take full advantage of this, we need to start on the right footing from day one, and we have the infrastructure to get all of our customers started that way.

[00:25:35] Ghazenfer Mansoor: Well, you talked earlier about the accuracy of the data. So obviously it’s very well known, there’s a lot of hallucination in AI. There’s a lot of… the AI would not always give you the right results.

Obviously it needs the right data. But how do you make sure that the results are that accurate, that the customers are excited with that?

[00:25:59] Harsha Rajasimha: So essentially, like with any technology, technology when used well will yield good results. When we use technology without understanding its strengths and weaknesses, it will yield poor results.

And so it’s one, understanding of the technology, two, following standard operating procedures and processes. It’s always people, process, and technology. It’s never technology alone.

I would be lying if I said we have an amazingly good AI based platform which never hallucinates and it’s always perfect. No, it’s not going to happen. We have to expect.

We have to understand the weaknesses of the technology that we are building and delivering, anticipate exceptions, have a way to handle those exceptions, enable humans with proper training, upskilling.

So we always say we have an AI ready platform. It’s not ready for all organizations. We need organizational readiness also.

It’s not just the technology readiness that matters, because not all organizations have people upskilled to understand the limitations of AI, to avoid misusing or incorrectly, unintentionally misusing the AI technology in a way it’s not meant to be.

And so it takes training, certification of people. It takes building standard operating procedures using this new versus old technology. If they were using paper and pencil, they were following a certain methodology and process.

Now, in some cases they were using legacy systems for electronic data capture. That was following a certain standard operating procedure. Now the same organizations have to be upgraded to prepare them to use AI based modern technology platforms, a clinical operating system that we deliver.

And that requires not just that our platform is fully validated, it also requires that our target users are trained and given the handbook with the right standard operating procedures that work in a regulated GXP environment.

And only when all things align, people, process, and technology, that’s when we say you’re ready for an AI driven clinical operating system.

[00:28:32] Ghazenfer Mansoor: Thanks. No, this is a really good insight. So I saw on your profile you started this nonprofit IndoUSrare. Can you talk more about it? What inspired…

[00:28:44] Harsha Rajasimha: Yeah.

[00:28:45] Ghazenfer Mansoor: …you, and what are the goals for that?

[00:28:48] Harsha Rajasimha: Absolutely, Ghazenfer. Thanks for asking. And that’s a 501 C three tax exempt nonprofit organization, also based here in Herndon, Virginia.

And largely based on my personal experience, as I mentioned, one of my three kids was born with a rare disease, and she only lived four days in the NICU. And I came across many families right here in the Washington DC area, but also back home in India where I grew up.

And then soon I realized globally there are many, many millions of patients and families affected by rare genetic diseases.

Certain communities have consanguineous marriages, marrying their first cousins, second cousins, and they tend to have more genetic diseases, like the Ashkenazi Jewish population is one example where we see the number of patients, families affected by genetic diseases.

And there are many, many other examples as well in the Middle East, in Africa, in India and other countries. And so as I realized that there is, of course I come from a technology background, but this is a human problem. This is a public health problem.

And many times these families have no hope. All they can curse is their fate or karma. And they think they did something wrong.

They have a sense of guilt. And technology alone cannot solve these types of problems.

And so what we built with the nonprofit foundation is bringing various people, families around policymakers, legislators, elected representatives from executive branches like FDA, NIH and others, as well as industry leaders, investors who invest in developing orphan drugs, clinical researchers, medical doctors, medical geneticists, genetic counselors. It takes a village.

And as I learn more and more about how some of the success stories that have come about, what are the patterns involved? What type of people and organizations and agencies do we need to bring together to make progress in ensuring that…

And that led to the vision that by 2050 we want to see all rare diseases treatable, and it’s doable with existing technology only if we can align the public health policy, drug development policies, healthcare policies, reimbursement policies are all aligned in a way that can make sure the weakest link in healthcare, which is rare disease.

If we can take care of the weakest link, we don’t need to worry about the strongest link in healthcare.

[00:31:47] Ghazenfer Mansoor: No, this is great. And thanks for doing that. This is a way of giving back and of whatever we achieve, giving back to the community.

That’s so great. Thank you for that. Okay.

[00:32:00] Harsha Rajasimha: You’re welcome. And we have on November first, we have a gala here at the Hylton Performing Arts Center in Manassas, Virginia, at the George Mason University campus. So I’m affiliate faculty at George Mason in the School of Systems Biology, where I do guest lectures in my spare time.

And so we have annual galas where we bring and honor families with rare diseases, but also people who have made a difference in global international collaborations for orphan drug development, genetics and medicine.

And so invite you and others who might be interested to get involved and join us at IndoUSrare.org, and you will find the information about the gala to join us.

[00:32:52] Ghazenfer Mansoor: So, yeah, we’ll make sure to include the link on our podcast description, and also we’ll make sure that we publish this episode before that time. Share the link. That way you can also…

[00:33:03] Harsha Rajasimha: Absolutely.

[00:33:04] Ghazenfer Mansoor: …share with our audience. So, yeah, thanks. So to wrap things up, there are a few, I’ll just do a little bit of a rapid fire round, small few questions just to change, I would say, the topic.

So what’s one habit that keeps you focused?

[00:33:22] Harsha Rajasimha: Walking, taking time, taking a step back and moving away from my busy work days, still for a half an hour to forty five minute walk every day. That’s one way, I would say. And also everyday morning routine, listening to certain music, slowing down in the morning before I start running.

[00:33:49] Ghazenfer Mansoor: No, this is good. And I read the other day that whenever you have anxiety or stress, the best thing is to move. Even if it’s changing a place, that’s the best thing to do.

Move, make a difference.

[00:34:02] Harsha Rajasimha: I don’t know.

[00:34:03] Ghazenfer Mansoor: What’s one thing clinical trials need more of?

[00:34:07] Harsha Rajasimha: Empathy, understanding of the other stakeholders. It takes multiple stakeholders, and each stakeholder thinks from their own perspective, naturally, but it’s very important to understand what 19 other stakeholders in clinical trials are going through as well.

What are their problems? What are they battling? And most importantly, what the patients’ expectations are and aligning around the patient.

[00:34:38] Ghazenfer Mansoor: What’s the best advice you receive as a founder?

[00:34:44] Harsha Rajasimha: Identify the product market fit before going too far.

[00:34:55] Ghazenfer Mansoor: What’s one book that influenced your thinking?

[00:34:58] Harsha Rajasimha: Steve Blank’s startup manual, which talks about the business model canvas. And get out of the building, go talk to customers, do the discovery interviews before you build a product that nobody wants. Right.

So you want to make sure you build a product that everybody wants, the target audience that you’re trying to help.

[00:35:18] Ghazenfer Mansoor: Yeah. That’s it. That’s his famous segment, get out of the building.

[00:35:22] Harsha Rajasimha: Yes.

[00:35:23] Ghazenfer Mansoor: Nobody will ever come between your couch and your TV to sell.

[00:35:26] Harsha Rajasimha: Mm-hmm.

[00:35:28] Ghazenfer Mansoor: That’s from a different book, but very similar. One word to describe the future of clinical research.

[00:35:37] Harsha Rajasimha: Unified and AI driven.

[00:35:41] Ghazenfer Mansoor: Cool, cool. Awesome. Thanks, Harsha. We have been talking to Harsha Rajasimha, who has shared such valuable insights with us today.

Harsha, where can people find you if they wanna talk more about the work that you’re doing?

[00:35:56] Harsha Rajasimha: Thank you, guys. And I’m easy to find on LinkedIn. I have a profile on Harsha Rajasimha. Please reach out to me on LinkedIn.

The two companies that I founded, Jeeva Clinical Trials. Jeeva Trials.com is the website. J E E V A Trials.com, and the nonprofit IndoUSrare dot org.

That’s also easy to find. And feel free to refer any patients who need assistance, and our team provides free assistance, care navigation service to families affected by rare disease.

[00:36:36] Ghazenfer Mansoor: Thank you. Yeah, we will include these links in the podcast description. So if you are listening, check out the podcast description and you’ll see those links and you just click from there.

Thanks again. So this episode is brought to you by Technology Rivers. At Technology Rivers, we bring innovation through AI and technology.

Most of our work is working with healthcare companies. First by helping businesses streamline their operations or building HIPAA compliant products. Thanks again, Harsha, for joining.

And thanks everyone for listening to the lessons from the leap. If you enjoyed today’s episode, don’t forget to subscribe and share it with someone who finds it valuable. Thanks again.

Thanks for making this an awesome show.

[00:37:18] Harsha Rajasimha: Thank you for having a very exciting conversation.