Revolutionizing Healthcare: Persistent Systems on AI, Digital Engineering, and Precision Medicine

Unlock the future of healthcare with our latest episode featuring Santos from Persistent Systems! Discover how this digital engineering powerhouse is revolutionizing patient care through enterprise modernization, AI, and data sciences. With a rich 34-year history, Persistent Systems has become a global leader in digital transformations, tackling complex challenges from drug development to hospital efficiency. Santos, an expert in clinical pharmacology and health IT, shares how his team's cutting-edge solutions are improving patient outcomes and advancing the healthcare ecosystem. Hear firsthand about the convergence of biological insights and AI technologies that position Persistent at the forefront of digital healthcare innovation.

Get ready for an eye-opening discussion on the role of generative AI in healthcare and life sciences. Santos takes us through Persistent’s unique engagement models and their rapid proof of concept initiatives like "Sure Start," which pave the way for transformative AI applications. From early diagnosis tools to precision medicine advancements in oncology, nephrology, and more, this episode underscores the immense potential of AI to revolutionize individualized patient care. Join us as we explore the dynamic advancements and continuous innovation within healthcare, emphasizing the transformative role of AI throughout the pandemic and beyond.

Transcript

Innovations in Healthcare and Life Sciences

Speaker 1 0:01

Hey everybody , fascinating discussion today , diving into the world of health , technology and life sciences innovation , with a true innovator in this space Santos from Persistent Systems . How are you ?

Speaker 2 0:16

I'm very good , Ivan . Good morning , I'm in Pune , India . This is Friday evening and it's my pleasure to be on this podcast with you .

Speaker 1 0:27

Well , thanks for being here . I'm really intrigued by the work you and the team are doing at Persistent Systems . Let's start with some introductions , not just to yourself and your background in so many areas , including clinical pharmacology and health IT , but also for those who aren't familiar . Who is Persistent Systems ?

Speaker 2 0:51

All right . So let me start first with Persistent Systems . Persistent Systems is a global digital engineering leader . It's in its 34th year after inception . Leader it's in its 34th year after inception . The company is based out of Pune , india , but now we have presence in about 20 countries . The total workforce as of today is close to 24,000 and we are close to a $1.3 billion company . The company predominantly excels in the area of digital innovations , digital transformation , partnerships . We serve multiple industry sectors , including the banking , finances , insurance , healthcare , life sciences , telecommunications and many more . Globally , we are recognized as a niche service provider in the area of digital transformations , digital innovations , and I particularly work at the interface of core domain technology , as well as all these digital innovations that we are talking about . So that's a little about persistent systems .

Speaker 2 2:01

I would like to tell you a little bit about my own journey .

Speaker 2 2:04

I started as a pharmacist , then I have done some previous training in the area of pharmaceutical sciences and then , for the past two decades , I have been working in various aspects of biomedical sciences , including the pharmaceutical industry , the med tech industry .

Speaker 2 2:23

I have been at Persistent for about three years . My current role I'm the chief domain expert for healthcare , life sciences innovations and in that capacity I work very closely with a multidisciplinary team of data scientists , domain experts , technologists , few people from the medical device regulatory industry . So it's a very vibrant , very interdisciplinary work culture and our remit is to look at emerging technologies that ultimately impact all the new things that are happening in the healthcare and life sciences domain . And particularly in the past two years , post-pandemic , we have seen a completely new landscape of healthcare innovations , perhaps driven by two different activities . One is the understanding that the entire healthcare operations is becoming more and more data-driven . So that realization came just out of the pandemic . And then in the last one year , all the big advances that have happened in the AI field with the advent of generative AI . So it's a very intriguing feeling to be at the interface of core domain and at the interface of this cutting edge technology , and I'm enjoying what I'm doing at Persistent .

Speaker 1 3:40

Wow , it's an exciting time to be in this industry and wow , I had no idea the scope and depth of your work at Persistent . And what kind of clients or challenges do you address in healthcare ? What's your mission there and your sort of unique position in the healthcare industry in particular ?

Speaker 2 4:03

Yeah , absolutely . Healthcare industry in particular ? Yeah , absolutely so . Like I said , the healthcare life sciences industry is undergoing significant evolution . Many of the big , large enterprises , they are adopting digital , modern digital technologies . So there is a big wave of what we call as digital engineering push . A lot of old systems are getting modernized . So that entire area is the area of what we call as enterprise modernization , where we need very good engineering and technology capabilities .

Speaker 2 4:37

The second wave of , I would say , technology adoption is at the level of data sciences in the healthcare industry . Adoption is at the level of data sciences in the healthcare industry . More and more it is evident that artificial intelligence be that your conventional artificial intelligence or the recent wave of generative AI it has tremendous applications in deriving actionable insights from all different types of data . So there is a very good use case , very good compelling reason for adoption of AI very early into any workflows . So that is the second wave of activities and the third wave of activities that there is a tremendous explosion in the understanding of life sciences , biology , medicine as such , right . So together there are three streams converging in the healthcare domain right now . One is this completely new way of looking at biology , medicine , synthetic biology and all these new things that are coming up . Then the advent of generative AI-like technologies . Maybe in future there could be another layer of technology like quantum . Very soon quantum is going to become a talking point . And the third is how are enterprises adopting themselves to embrace these kind of technologies ? So these are some industry challenges we are facing and this is exactly the area where Persistent has excelled .

Speaker 2 5:59

In the healthcare life sciences arm of Persistent , we serve a wide spectrum of clients all the way , starting from scientific instruments , medical devices , diagnostics , pharma companies , contract research organizations , hospitals , insurance payers , providers and everybody in the ecosystem . So in some way Persistent has differentiated ourselves by calling ourselves as an ecosystem service partner to serve the entire needs of the healthcare ecosystem . If you really think about the ecosystem , there are no silos here because ultimately you are solving a problem and that problem is ultimately going to benefit the patient at the end . So these different sub-segments of the industry ultimately converge upon . How can you improve the final outcome from a patient point of view and things like that . So Persistent has this motto of talking about cell to cure journeys , where we take note of all the cell biology , the molecular biology advances and then convert all the way into digital care pathways , a lot of personalization of medical concepts , and this entire journey is now getting digitized . So in that space , we are considered as one of the leaders in digital innovations for healthcare , as well as for the pharma industry .

Speaker 1 7:27

Well , fantastic mission , great work . And I'm based here in Boston and we do some things very well things like patient care and life sciences , and biology and biotech and on and on but some things , in my opinion , not as well cutting edge IT practices and data science and some of the execution of those technologies . So , when it comes to AI and IT and related topics , I mean , what work do you specifically do with clients ? I mean , how do you help ?

Speaker 2 8:04

So , with regards to applications of AI in healthcare life sciences I'm using the term healthcare , life sciences because that is an industry term , but if you really think very hard on this , these are integrated words . Life sciences percolates into healthcare and healthcare depends on life sciences . How do we help our customers ? First , the entire journey of what we call as getting ready for the latest kind of data practices , right ? So we all know that unless you have a good quality data , your downstream ai applications are moot . So the very first support that we provide to our customers is to help them understand their current data practices . So how can you create better data platforms ? How can you help curate the available data ? How can we introduce some quality control measures in the available data ? How can we manage this data very well ? So this entire first piece of work is what is called as data engineering . And , having established some good , organized systems for handling healthcare data , then we take them into different kinds of downstream applications , which are then ultimately powered by either your conventional AI or generative AI . Now , in this space , just to give you a few examples , we have worked very closely with large companies as well as startups in order to develop some AI-based products , just to give you a few examples we have worked in early detection of lung cancer from liquid biopsy samples , and the idea here is that when you take a blood sample , you see large number of normal cells with a small number of cancer cells , and those cancer cells have a different morphology and therefore they have different patterns when you observe them under a microscope . So you can train a model to differentiate a cancer cell from a non-cancer cell , and that can be done with the help of AI . So that's one example . Then we work with a few companies who are in the space of early detection of kidney diseases . So by applying AI principles on available clinical data , you can stratify individuals based on high risk category and then low risk category , and then obviously people in the high risk category can be looked at with a slightly different lens of medical treatment . So that is at the level of , let's say , screening and early detection .

Speaker 2 10:26

We also work with a lot of our clients to develop disease monitoring strategies . So , for example , if you are working with cancer patients , how do you monitor the patient with the help of almost data that is collected daily or weekly , and how can you track the patients ? How can you do longitudinal data collections . How can you do predictive analytics to determine what may happen to this patient two years down the road ? So all this area of work is in the area of therapy monitoring . And then recently we have also started working with our clients in the area of population screening , and population screening is becoming a very popular area right now because the philosophy is now moving away from disease but into wellness . The philosophy is now moving away from disease but into wellness . So how can we bring wellness thinking so that we can slow down the processes of disease creations ? So in this place we try to understand what are the determinants of good health . Can you apply some AI principles to find out what may happen two years down the road and try to bring in early interventions right ? So this is in the area of diagnostics . Taking a step further , we work very closely with pharma companies where we help them create drug discovery platforms , help in clinical trial designs . We try to help them match patients to a particular requirement of a clinical trial . And this is a space where we work very closely with CROs , contract research organizations to bring in a lot of AI capabilities into their current product pipelines .

Speaker 2 11:55

Taking a little further about patient management , so how can you serve a patient better , all the way from hospital admission , the processes during the hospital , even monitoring at home ? So the new concept in patient care is that most of the patients now can be managed at home so it's called as a hospital at home concept and in that you need a lot of wearable devices , and those wearable devices generate data 24-7 and that data has to be put on a cloud and through that cloud you have to then do some kind of data analytics . So that is how we work with a lot of our healthcare partners to do these hospital at home kind of services .

Speaker 2 12:35

And lastly , in the insurance sector , there is a significant impact . Unfortunately , many patients do not get their insurance claims honored because of lack of data or maybe some kind of a knowledge gap . So can you create systems which are driven by very good data quality coming from the electronic medical records and then apply some AI principles by which you can honor the insurance claims . So that is a very high impact area of work , because nobody wants to not give a medical coverage to a deserving patient . So this whole area of using insurance claims intelligence is also significantly bolstered by the application of AI products , right ? So I think AI has a logical place in the entire spectrum of healthcare life sciences . It is just . The requirement is that you have to have very good data systems to maintain good quality . And then the logical choice of AI , whether it is predictive AI or generative AI .

Speaker 1 13:40

So we work very closely with our clients to do this kind of strategic thinking and then implementation of such strategies . Well done , congratulations . That's impressive . Let's shift here to generative AI . Lots of opportunities there on the horizon . Of course , this is much more than just using chat , gpt or creating funny pictures through Dolly . There are real practical use cases here for generative AI and precision medicine and beyond . Maybe throw some light on what you're doing in this space and the opportunities that are out there .

Speaker 2 14:11

Yeah , absolutely so , as you said , I think . Unfortunately , generative AI , definitely in the last six to eight months , has created a lot of buzz , but there is also a little bit of a hype cycle in generative AI , and we are seeing that in healthcare life sciences , perhaps also driven by incomplete knowledge about the applications of generative AI , and we are seeing that in healthcare life sciences , perhaps also driven by incomplete knowledge about the applications of generative AI . Let me take it maybe step-by-step . In general , ai , or generative AI , can always be positioned as an assistive technology . It is never going to be a technology that will obviate or it will surpass a human expert . So the entire positioning right now that is happening within the industry and even with our clients , is that how can we improve the efficiency of human experts with the help of AI ? So it's very clear that , first of all , it's not an absolute technology , but it is an assistive technology . Now , having said that , how do you bring the generative AI technology adoption into various workflows ? So , at Persistent , what we have done is we have created a generative AI task force . It started very early , into January 2023 . And this task force started looking at all the trends that were coming up in the healthcare life sciences market .

Speaker 2 15:29

In doing that , we also decided internally that we need to do two or three things . The first thing is that we wanted to train our workforce . Um , so , out of the 24 000 uh colleagues at persistent uh , almost 60 to 70 percent underwent generative ai training so that when you are facing clients , uh , you are capable of handling generative ai like workflows . So there was an internal push for upskilling . The second is , uh , the story of generative AI is the story also of hyperscalers , right . So these are cloud services , because delivering a generative AI like solution requires high computing power and for which you need cloud services and all the major clouds that are currently dominating the market . We establish strategic relationships with all of them these are called as strategic partnership agreements by which we get early access to their products , on which we can do innovations , and then we can show our clients some value propositions . In addition to that , persistent engineers also contribute to the innovation index of those hyperscalers right , so it's a bi-directional relationship to the innovation index of those hyperscalers , right , so it's a bi-directional relationship

Implementing Generative AI in Healthcare

Speaker 2 16:43

.

Speaker 2 16:43

The third thing that we did was we started working very closely with our clients in creating industry-specific solutions . So , for example , my own sector , which is healthcare , life sciences , we started creating our internal intellectual property , as we call them accelerators . And these accelerators would imagine that there is a real-life problem , let's say , in early diagnosis , and how can you use generative AI to assist early diagnosis ? Anticipating these industry problems , we started creating our own IP pipeline . This IP pipeline , then , would be shown to our potential clients that look , this is a proof of concept that we can solve a certain problem with the help of this approach . Now , these accelerators are minimum viable products . These are not finished products as such , but by showing these capabilities we generate a lot of client confidence that Persistent can therefore deliver these complex technologies .

Speaker 2 17:35

And lastly , generative AI being such a dynamic field , we created some unique engagement models , unlike your regular technologies , generative AI . Nobody wanted to go full hog on generative AI because that was an evolving piece of technology . So we created some unique models of engagements , which were like eight to 10 week models we call them sure start program and during that program we actually create small POCs . Pocs are proofs of proof of concepts , and during that period we are able to demonstrate that a very complex problem , let's say , choosing a drug for a cancer patient , can be demonstrated with the help of a generative AI technology . The advantage of this eight week period period is that it is a fail-fast , fail-safe program . So , for example , if the technology is not working for the given problem , you can actually pull the plug fairly early into the game right you cut your losses . On the other hand , if the proof of concept is working very well , you can then do the entire scale-up planning and then you have a proof of concept through your demonstration that this can work , and then you can go back to the board and then you can do scale-up production of that entire process right . So these are four sort of major activities that we undertook training , doing hyperscaler partnerships , then doing internal solutioning , creating our own IP and then creating unique business models , and thankfully , this worked very well for us at Persistent .

Speaker 2 19:06

Persistent is being recognized by a lot of industry analysts as one of the leaders in the delivery of generative AI solutions , and our CEO , mr Sandeep Kalra , and our founder , dr Anandesh Pandey , have called it in public fora that Persistent is now being recognized as an AI-first organization . The entire philosophy of adopting AI right from the outset has been imbibed into the entire ecosystem of Persistent . Now we are taking a very focused AI-based approach . Generative AI is now being used for two purposes . One is AI for engineers that means building solutions and the second is AI for business solutions that means to create unique use cases , unique propositions , unique platforms , unique co-invention models , etc . So it's a very vibrant area of work right now and we actually are being recognized by the industry , which gives us some gratification that this is one area where we would like to play for a long time to go in near future . So that's our generative AI story so far .

Speaker 1 20:12

Wow , very comprehensive , very impressive . I had no idea you were at the cutting edge of what's happening in this space and yet there's a lot of skepticism , reluctance . We're a pretty conservative group in healthcare , for good reason . What are some of the challenges and roadblocks that need to be overcome to really implement Gen AI in medicine in a secure , practical , reliable , compliant way , you know , to really achieve these opportunities , what do you think ?

Speaker 2 20:44

Yeah , absolutely . Unfortunately , the society perceives generative AI as a magic bullet , right ? If you look at the way the media covers generative AI , it's almost like you know you're going to be in a completely new era of medicine , et cetera . I think all those are good things when you look at the rhetoric of this whole process , but there is more than meets to the eye here . So , first , all the generative AI models that we currently work with I'm not going to name any one of them , I'm just probably taking like a neutral view . Many of them are built on a lot of data that has come from different healthcare sectors , but there is an inherent bias in this data , right ? So that data may be lopsided for , let's say , a certain population or a certain geography , or maybe only certain type of disease areas , et cetera , right ? So whatever data has gone into these generative AI models is not the data universe , but it is a slice of the data universe , right ? So that is a fundamental limitation of the models that we are talking about . That's one . Secondly , as I mentioned earlier , that generative AI , or any other area , at best can be considered as an assistive technology , so we have to remember that it's not going to be a hundred percent accurate technology . It is an evolving concept . There are some limitations in the current models . We need to put some guardrails in our own thinking of their applications and ultimately we have to see that what is the difference between a signal to noise when it comes to applications of this technology ? So that is the second aspect . So we need a lot of awareness about the limitations and the strengths of such technologies . That is the second aspect . The third is these technologies are constantly evolving Right . So somewhere in the literature there is a very interesting mention that the half life of a generative AI concept today is about three to four weeks , so you can imagine the turnover time . It could be even less looking at the way information is released in the public domain right now . So this is a very , very fast moving , very , I would say , amorphous structure . Right now , that means a lot of new things are coming , so you need to constantly track what is happening in the area . There is not going to means a lot of new things are coming , so you need to constantly track what is happening in the area . There is not going to be a rule of thumb here . It's going to be a constantly evolving space , so the last word has not been said on this particular technology . You will see a lot of this , new releases that are coming up together , so the ultimate challenge here is , considering that this is such a dynamic space , how do you make yourself ready to play in this game , right ? So that is what everybody , at an individual level as well as an organization level . They are constantly thinking about this and everybody is trying to find a niche area of differentiation , and I would like to talk about one of our niche area solutions . We have recently created an accelerator , which is called Generative AI for Precision Medicine .

Speaker 2 23:39

So what is precision medicine in today's world ? Again , it's a buzzword . Precision medicine is a new model of looking at patient care where , unlike in the past , when all patients were treated equally , now patients are treated with the help of individualized paradigms . So what I mean by that ? There are unique biomarkers associated at the genetic level , at the protein level , at the imaging level . Let's say , my biomarkers are different than your biomarkers , and that makes us uniquely different . So whatever applies to me as part of the treatment may not apply to you , right ? So that is the realization that has really sunk in into the system and therefore we now need to do an individualized approach , which is called as precision medicine .

Speaker 2 24:25

Now , in this space of precision medicine , we have created our end-to-end comprehensive solution , which is called as a platform for precision medicine with generative AI , and the platform has about four , five modules . The first module is when a patient walks into a hospital and interacts with a patient , interacts with a physician . There is a lot of audiovisual data . So , for example , the way we are speaking right now , we are generating a lot of audiovisual data , but in that data there is a lot of content , right . So you can convert this audiovisual data into a transcript and from that transcript , which is unstructured data , then you can apply generative AI principles and then you can create structured data that becomes the intake into electronic medical records . So currently the problem is doctor is not able to give time to the patient to do eye contact during the interview process . But when we do this kind of a system , we can actually have a better physician patient contact and more engagement , right ? So that is model number one .

Speaker 2 25:24

Model number two is almost in every disease area today , there is a tremendous need for doing advanced imaging , and that advanced imaging could be at the level of the cell . It could be at the level of the tissue , it could be at the level of the organ or the whole body , right ? So there are concepts in pathology , there are concepts in radiology , where you need advanced imaging , and in this space there are some new models that have come up . These are called as large multimodal models , lmms and these LMMs have the unique ability to read an image and then give you medically sane output right , so it's called as an image-to-text conversion and then give you medically sane output right , so it's called as an image-to-text conversion . And one can imagine this system can be utilized for reading an image and creating a first draft of a medical report and please remember , it's an assistive technology , it is not a final draft but the process of reading the image can therefore be automated and that will significantly help a doctor do better report generation and ultimately it will benefit the patient because your turnaround time for report generation can be significantly reduced . So that is the second module that we have created , which is called as image to text analysis , with the help of large multimodal models . The third is now the doctor needs to create automated reports based on the data that is made available . So , using the concepts of LLMs large language models . We have built systems by which we can generate reports automatically and in this process what we do is we take key elements of the crude report and then we auto-populate into a standardized template which looks like a final medical report . So that is module number three .

Speaker 2 26:57

Module number four is the medical data is exploding daily . There are a huge amount of medical papers , a lot of clinical guidelines that are coming up , and it is impossible for the doctor to track all this on a daily basis . So here we have created a generative AI-based solution . It is based on a concept called as knowledge synthesis , where you can look at large amount of medical literature and you can break down into small segments on which you can do very , very specific querying with the help of a virtual assistant . And to ensure that the virtual assistant does not have any medical errors , we regulate the output of that virtual assistant with the help of a technology called retrieval augmented generation , rag . So ultimately , when we query with the help of a virtual assistant , whatever answers we get are medically authentic , with citations right . So it's a very important parameter for a doctor to get authentic information at the tip of your fingertips right , and it's on in real time . So that is a process we have generated for doctors to look at medical literature .

Speaker 2 28:01

And lastly , when a doctor needs to decide whether to give drug A or drug B , a doctor uses algorithms , which are called as evidence-based medicine , and evidence-based medicine is based on two major decision-making levers . The first lever is whether this drug has been proven with the help of clinical trials , and the second decision-making lever is whether this drug is going to have any toxicity to important organs in the body . These are the two most important parameters on which a doctor will decide whether to give a drug or not . These are the two most important parameters on which a doctor will decide whether to give a drug or not . And all this decision making is based on some kind of a medical algorithm , and we have created a solution by which we can train a model to facilitate this kind of medical decision making that will ultimately help a doctor choose the right drug , given a profile based on , let's say , genetic testing or pathological testing , etc .

Advancing Precision Medicine With AI

Speaker 2 28:52

Right , and the last and the most important model is the insurance sector . Now medical costs are skyrocketing , so anybody without insurance sector virtually has no access to medical treatment . But for those people who have access to medical treatment . The rate of insurance claim rejection is very high and for that reason , using a concept called as cross-matching of medical terminologies with our MediClaim document in United States , we have been able to show whether a particular patient's case can be honored with the help of medical insurance claims or not . Right . So this whole process of six modules that I just talked about , individualized to the needs of the patient based on the data coming from their electronic medical records , is at the core of precision medicine , and for each of these processes , we have applied either a large language model or a large multimodal model , or a combination of both , to show that these processes can be enhanced .

Speaker 2 29:48

The good news here for us is that all of these solutions are no longer lip service or not in theoretical mode , but we have started demonstrating the value of this to many of our clients .

Speaker 2 29:59

These clients are taking them as either a point solution or maybe like a combination of three or four solutions to implement into their ecosystem . We are seeing more and more adoption of comprehensive applications , of generative AI not point solutions , but comprehensive applications across the landscape , and I think that is one area where we will continue to grow as an innovator in this area , where we will keep looking at innovating in precision medicine , let's say , with different disease areas . The key disease areas where we are seeing impact is oncology , definitely nephrology , kidney diseases , cardiovascular diseases and diabetology . So these are few , I would say , focus or thrust areas that Persistent has identified and we are constantly trying to look , breaking new horizons in these particular areas right . So I think this is a very ambitious , very challenging space right now , but looking at the early success that we are getting out of this , I think we are fairly confident that we can probably make a dent here .

Speaker 1 30:57

Well , make a dent , to say the least . Congratulations on that incredible work . I had no idea the amount of innovation investment you've been putting into this space . Well done . Just on a final note here personally , you've dedicated your professional life , personal life , to healthcare and medicine . What are you excited about most personally when it comes to our future with AI ?

Speaker 2 31:24

Oh , I think the big revelation for me being in this area was during the , the second and the third wave of the pandemic , and one could see that the entire global community was coming together thanks to technology . There were people , let's say I was connected with people in Brazil . I have never met so the advent of advanced technology and the ability to take a technology concept across borders and then helping each other out on a daily basis , challenging status quo , constantly looking at innovations , trying to unlearn what we have learned in the last few years and try to learn new concepts , I think these are all multiple triggers . You know , like I said , after two decades in this industry , I'm very excited . Like I said , after two decades in this industry , I'm very excited .

Speaker 2 32:12

It almost feels like you're like a kid in the candy store looking at all the new things that are coming up . And I think this is like I said the last word has not been said yet . I think the best is yet to come . There are a lot of new ideas , new inventions . There are going to be another technology waves coming every once in a while and I think this is where we need to constantly look at . So it keeps you fresh , it keeps you proactive , it keeps you honest , and I think that is something I really love about what is happening in this field right now , and I'm almost seeing like a second career wave , even for myself .

Speaker 1 32:47

Well , congratulations on that and great work onwards and upwards . Appreciate your time and all the insights and reach out everyone . Follow Persistent Systems on social media and check out their website . I really enjoyed your blog Really informative , educational , very practical , insightful pieces there . So thanks , thanks everyone , thanks for listening and until next time , take care . Thanks everyone , thanks for listening and until next time take care .

Speaker 2 33:15

Thanks , ivan , for having me here . Thank you , thanks . Bye-bye .