Generative AI: Insights On Advancements 

Speaker A: Hello world. This is Better Tech, a ah, podcast where we chat with some of the most successful leaders about the latest industry developments. So join us as we explore the world reliance on tech.

Speaker B: Hello and welcome to Better Tech. I'm um, Colin McCarthy, your host. Here today talking to Indradit, the CTO of cellstrat, the makers of Cellverse AI. Welcome to the show. Before we get started into a discussion about servers AI, can you briefly introduce yourself, give uh, myself and our listeners a bit of background into your journey into AI and uh, what led you into this space.

Speaker C: Hi Gaulin, very good morning and thanks for having me here. Uh, so my name is Indrajit Singh and I am the CTO and senior generative AI architect, uh, at salestrack and I have been leading a lot of uh, solutions at sales track. So my work primary involves developing innovative and cutting edge solutions in generative AI and computer vision. So my journey started um, you know, in 2010 when I was in school when I actually got more interested in coding and programming and databases and all those stuff. Uh then I joined some of the companies and product based companies like etl. I worked on big ah, data, data warehousing and those stuff. Uh, then from there I continued to the data space and AI and uh, mostly NLP where I got more interest. So that's where it began and I got more interested in generative AI.

Speaker B: Right, right. Yeah, you've been on that path for sort of the last 10, 15 years where AI has really come into everybody's uh, daily life. Uh, what actually inspired the creation of uh, Silverse AI? Um, what was the particular problem or market gap that drove the product to be created?

Speaker C: Yeah, so uh, Salverse is really an interesting um, journey for all of us at Salstrat. So we defined three core uh, problems that were consistently um, hampering the progress of the medical researchers. So basically they are bogged down by a lot of uh, severe time constraints. They are really, really um, having very, very um, tight deadlines. They're running on their information, uh, is also having a huge overload like the insights. They are uh, often buried, um, uh, and the instructions that they are given to search for right information is really uh, difficult for them to cope up. So one of the thing is that the severe time constraint and information overload. And the other thing is that we see uh, outdated tool sets. Right. Especially in India as a geography, uh, we see that the medical researchers are often using old tools. They are uh, far behind in the technology. Maybe they're not uh, equipped with or they do not have the time to even look for certain tools. So that's where we actually thought of helping these medical researchers to come up with some uh, advanced tools, especially generative AI powered so the search becomes easier for them.

Speaker B: Right, right. It's an amazing use of AI in a, in a specific field with you know a uh, very well trained and specifically trained uh LLM large language model to help a specific market. AI is great but we do need these special niche uh, highly trained models to really uh, empower those uh, industries get the most out of it. And it is a wonderful tool as you say for that uh, research. Can you share some sort of real world examples where Silverse has uh, delivered some specific impact, you know, across all of the places that it's being used. Healthcare, life sciences, et cetera.

Speaker C: Yeah. So we have basically demonstrated this product not only in India but as well as Middle east and Europe. And our platform has automated the analysis of thousands of documents allowing the researchers, researcher Personas specifically as we built it for medical researchers specifically. And they are able to ask complex questions like you know, let's say for, for example what are the imaging uh gene, uh targets for um, you know, autoimmune disorder, um, with some known safety profile and you know, preclinical studies. So um, the, the big difference uh, from ChatGPT and other tools, I mean I would not uh, really particularly name some of these tools but you know they are basically very um, generic. They are searching everything from the Internet and you cannot really rely on that information. Um, while Silvers is more focused towards specific journals and very authentic source of information, uh, and that's basically giving them more confidence that they are not really searching unnecessary information. One is the time and the other thing is the authenticity of the information. So the researchers in Middle east, uh, some of the pharmaceutical companies where we are actually already deployed this product uh as a white label cell bot and we also deployed this in Bangalore and in India and Delhi. Some of the medical uh, research um, hospitals ah, where the people are using it and there's a phenomenon feedback for it.

Speaker B: Right? Yeah, it's you know those other models as you mentioned will do have a very creative element, are ah, able to do hallucinations and we rely on hallucinations for that creativity. When you're doing research you need it to be grounded in the data that you have given it and only provide responses based on its known data.

Speaker C: Absolutely.

Speaker B: Um, and this is why uh, uh, any company's journey, regardless of whether they're doing a medical research uh, AI platform Like yourselves or they're trying to utilize AI in their business. I think they have to have um, a lot of confidence in their own company. Data, um, and the data and what the model is learning on is really the key aspect here. Um, the one thing that I like about AI platforms and I do use Gemini an awful lot in my daily life and my working life is the way that we can absorb data. Um you know with NotebookLM we can create audio overviews. With Gemini we can create um, some very nice infographics to sort of digest uh, and view that information. Silver sort of is standing out in the market for combining uh gen AI with immersive technologies which is a great way for medical professionals and researchers to view and analyze that data. Uh how does this integration enhance the sort of the medical researchers experience and their outcomes from what they're learning? What are you hearing from them?

Speaker C: So some of the you know the feedback that we have received from these researchers is like they are telling this tool is very useful in terms of getting information and that only concerns by medical field. And it is also incredible website for some of the uh, you know folks in the healthcare field. Like they think it's better uh future for them as a medical student uh because the uh finding authentic information was like getting abstract right the way so and it also helped them figure out uh, figure out some of the information that are really crucial in terms of not really getting from uh other tools and Google search. Right. So and also they can upload their own research uh um you know PDF files or research articles uh and they can chat with it and they can come up with a uh better flow in their research. These are some of the very useful features and feedbacks that we got.

Speaker B: Right. Yeah, I love the ability to do that um In Gemini's notebook lm upload my own PDFs Ah and then chat with them and get a response directly out of them. It's great for your medical researchers to have the same experience in Salverse. You know obviously this is AI is, is transforming healthcare training. Um, in what ways is self us actually revolutionizing this training um and the simulation for healthcare professionals.

Speaker C: Yeah. So I think one of the things that I mentioned about the time and the speed. Right. So for the, for India specifically the researchers are not really focused on Internet based search or you know the things that they are not relying on. So the confidence and the belief, you know that matters a lot for uh, you know in this kind of field. So what we are trying to do is we try to connect with each of these researchers personally and understand their pain points on a day to day basis and try to solve those pain points for those researchers. And they have really given us a very good feedback and we're improving the tool belong back.

Speaker B: Yeah, right, yeah, it's, it's great to hear you talk about finding those pain points um, because I think there's a lot of misconception that AI can do everything but it's, it's the. No it can't. It is there as, as a productivity enhancement, not a productivity replacement tool. Um, you know you, you, you still have to do the work. It's not going to replace everything. It just makes it a lot easier. Um, um, it's almost like having you know a very highly skilled intern or a colleague, you know a coworker, a co editor, um, a collaborator, uh, and can automate some of those repetitive tasks. But you know, you still need to have human in the loop, um, confirming that everything that is being generated is accurate and having the final say. Any AI rollout and deployment is about finding those pain points. And I think even you know, in business particularly as well as sort of the medical field, it's looking at the processes, daily tasks, the structure of the work and then re engineering it with AI in mind, fully utilizing those tools. One of the problems you've probably seen, and I know I've certainly seen in business is helping to drive adoption. What are the uh, biggest challenges for onboarding new users to these immersive platforms and how does sellvos make that process a little bit easier?

Speaker C: Yeah clarinet. I think you have hit the central challenge, you know the biggest barrier to adoption and it's basically intense demand on the time, uh and attention of the healthcare and you know, the research professionals. So they are you know, rightfully skeptical um of any new tool, uh technology that is being coming in the market. But the thought process is changing and we have been working with a lot of hospitals and medical researchers directly talking to them and I think the thought process is changing and they are also trying to adopt AI as an, not as a replacement as you said rightly, uh, but as an assistance but as a, as a helpful tool to um, make their uh, journey much more easier the time, uh, and focus, uh, you know, sensitive. So we are actually making it more intuitive uh like and giving in conversational interface like sometimes the researchers are ready so really busy so they can just have a voice command, you know, hey can you search this for me and quickly tell me what should I do for the, as a next step. Right. So that Helps them quickly. I mean instead of like chatting and you know, uploading a PDF so you just do a voice command. Hey, can you upload this PDF search in Google Drive and it will search that PDF and then upload that uh, in the Cellbot, you know, research tool. And then their research is ready pretty much. Right. So when you are like too busy and so, and also we are making a zero friction onboarding like you, you're just having on web based platform. Right. So you're requiring no complex installation it so overhead and so on. Right. So uh, in a medical setup it's very difficult for them. I mean they don't even check their emails. I mean to be, I mean in India we have seen that you know sometimes there are two um, busy. Ah so yeah, so these are the things that we have taken as a measure. Uh, and we see that the adoption is far um, faster and we also go to the, to the, to this, some of these colleges talk to them with medical researchers and also ask them how they like it, what is the experience and if they wanted to use the tool. Right. And that is definitely helping us then share our tool with other medical researchers.

Speaker B: You talk about the difficulty of, of people even looking at their own email. You know it is, we know that 89%, I think I saw one study, 89% of corporate users have to look in like six different places for their, for their information and then you know trying to give them another, another tool, another tab to look at, um, can be challenging. Um, but if that, yeah if that new tool or that new browser tab brings in all of the information that you need and gives you a single point to do that analysis uh, across all of your data stores, then ah, ah, it's a winning solution. Naturally it is in the browser. Uh, nobody should be installing applications nowadays. Uh, with zero trust it's so easy to just have everything in the browser and it does give a much better sort of almost a more secure experience I think.

Speaker C: Yep, yep, absolutely.

Speaker B: Um, talking about security, uh, privacy and compliance, uh, obviously in healthcare, you know in the US there's, there's hipaa, other countries have, have other um, requirements Regulatory uh, compliance for healthcare data. How does Silvers ensure the security, uh, and all of those global regulatory compliances when you're working with very, very sensitive uh, uh data sets?

Speaker C: Yeah, very very important question Colin. So data security and regulatory compliance are foundational to any platform architecture, be it healthcare or any other domain. Um ah, Salverse currently is designed to work primarily with public and anonymized data Sets like from sources like scientific literatures, clinical trial data that are available to everyone. But by focusing on these kind of type of information, we intentionally avoid processing PII or PHI information. But we have a clear roadmap to uh, incorporate sensitive data analysis as we kind of evolve uh, in this space. You know, so the researchers are, or may they may uh, also going to use some of these information that are PII sensitive and we need to remove those PII information early on before we even push it to, you know, cloud.

Speaker B: Right, right. Yeah. Uh, I'm just thinking of the, you know, the security of everybody's pii. Personally identify, personally identifiable information is very important. So yeah, it's great that you have that removed so that the researchers, those publicly accessible um, data sets, um, obviously I'm sure security is something that you've had to come across in your career history as a cto. Um, one thing I would love to understand and ask you a question about other leaders. Um, certainly in this new emerging world of AI, what's the sort of the one quality that you think, uh, differentiates at the moment between good leaders and great leaders in the tech space?

Speaker C: I think Colleen, empathy is one of the major uh, you know, qualities of a good leader. Uh, if somebody is not able to uh, perform a task instead of you know, going into a problem that that person has, go to the root cause, maybe give some time. Is that needed learning, they need a training. Provide the training, empower that person to do 10x than what they were doing instead of pressurizing and you know, making it a fish.

Speaker B: Right, right. Yeah. Um, we've talked a lot about uh, healthcare, uh, Selva Self Strat company you work for. Want uh, to ask you where you think uh, AI is going to completely transform another industry. You know, uh, Salvos. AI is having a great impact on healthc there. You know, you've got wide adoption in India. Where do you think uh, AI can also um, be. Be very transformative?

Speaker C: Yeah, I think Colin, I think each and every industry, but as you asked very specifically I believe sports, uh, retail industries especially and the complete real estate, um, you know, think about real estate search. Colin. Um, how do you, how do you search properties? Uh, you go to the site and then you probably look for certain apartments. Um, if uh, you have to look for rent or buy, uh, think about searching globally and you want to invest in some properties that are within some range and it can search global sites and gives you the list of top properties that is, you know, relevant to you within your budget and the legal Litigations and other stuff, you know.

Speaker B: Yeah, that's, that has really got me thinking because ah, I have in the last few months moved, uh, I moved from New Jersey to Maine and I spent a lot of time on those uh, real estate realtor sites. And we would generally search by the size of the property, the number of rooms that we wanted and uh, our uh, price range and we had to go through and we had some other key things that we were looking for but they weren't generally in the description or the selectable search. So if I could be descriptive in a large language model and describe what I, what I wanted, you know, then, then the experience of uh, houses being you know presented to me might be more relative. They might not meet that, that drop down, you know, checkbox, uh search that we have at the moment, but their you know, description and other elements of that, that building might be more to my liking because I uh, you know of how we've been able to describe what we want um, as opposed to have a very traditional search functionality. So yeah, that's interesting how, how that will revolutionize that market as well.

Speaker C: So Echelon I want to add within with that is basically the agentic AI.

Speaker B: Right.

Speaker C: So how agentic AI is helping in these kind of search, I mean just to take an example of these real estate search agentic, the LLMs are not good at you know, searching because they are trained with uh, enormous data and not recent data. So if you search like what's going to happen tomorrow? It might say like I'm not sure what's the weather today? Right. It might not give you the answer. What if a tool that you know, allows these LLMs to connect to weather, connect to real estate sites, um, the, the, the, the current information that is available to particular state, particular city, particular country and searching that information and presenting you with uh, precise information that you are looking for. So agents are doing that?

Speaker B: Yeah, yeah, yeah, yeah. Agentic AI is going to continue to be uh, as transformative as AI has in the last couple of years. Um, and I'm always reminded that AI didn't just magically appear two years ago when OpenAI announced ChatGPT. It's actually been there for quite a while. Um, and Google and other companies have been pioneers um, for the last decade, um, uh, looking ahead at uh, the future, um, what's next for cellverse, uh, in the coming years, uh, where do you see spatial uh, computing and immersive AI heading?

Speaker C: So I believe the future is about um, not just looking at charts or textual information Colin. Right. It's about walking through a 3D model or AR VR, a system where humans or brain, or let's say manipulating a protein structure is shown to the, to the researchers and to doctors, you know, and they interact with it. Not just being shown, they can interact uh, with an capability of finger or eye movement. Right. So um, to find new drug compound or new disease or a new treatment plan. Right. So imagine a team of researchers, uh, from across the world, right. Uh, across the globe meeting in a shared virtual lab or like notebook or a kind of platform. Right? So that visualize how um, a virus attack a cell, uh, in a real time guided by AI that highlights the key interactions and suggestions, Right?

Speaker B: Yeah, I think it is really uh, going to remove those barriers, those borders and make uh, global collaboration, teamwork, research, uh, just so much better. Um, it's been great to talk to you for the last sort of 20min. Understand uh, cell strat, your, your view of, of AI in healthcare and learn about cell, uh, verse AI. The one thing I did like, uh, about the site, uh, Salverse AI is, is there is a free tier. You know, you have your pricing there.

Speaker C: Yes.

Speaker B: If companies are interested, people can uh, log in with their own Google account identity through owas, uh, and look at the platform themselves and do some analysis, um, and see what they get. Yep. Well, thanks once again this has been a, uh, better tech podcast.

Speaker A: We look forward to bringing you the latest industry news in our next episode. In the meantime, check out our other episodes@texcel.com podcast and be sure to subscribe to our YouTube channel so that you never miss an episode.