Disambiguation: Generative AI for Life Sciences - Transcript

Michael Fauscette

Welcome to disambiguation. I'm your host, Michael Fauscette. Each week we interview experts in AI generative AI and business automation to help business leaders understand how to use these tools for the biggest business impact. In our show today, we look at the use of generative AI and Life Sciences. I'm joined by Michelle Wu, CEO and co-founder of NyquistAI. Welcome. Hi, Michael, could you could you give us a quick introduction and talk a little bit about what you guys are doing it at NyquistAI.

Michelle Wu

Sure, sure. Absolutely. I'm the CEO and co-founder of Nyquist, AI, and big nerd, all my life has been in life science. And how I came up with the idea of Nyquist AI is to my real-life experience, I was the youngest Global Strategy Manager at Novartis and worked on the first the very first and so far the only asset swap deal in the history of pharma, like three way merger and acquisition among portfolio GSK, Eli Lilly, and Novartis, which is a lot of data. And after burning so many late-night oils, and bending my head against the wall and calling by expert friends, like what does this document mean? Or what does this clinical trial report me and then I thought like, well, there should be an easier way to analyze, like, first aggregate all the useful information in life science, and then make sense of those data and, you know, make actions and decisions out of this data in a much like painless way. So that's how I came up with the idea of Nyquist AI. And so our company aggregate, all the global clinical trial, regulatory documents, post market surveillance, like the adverse event data from around the world, we are approaching our third year anniversary at Dell, we have built the largest training set globally, from clinical trials in the US, in China, in Japan, around the world. So what is really serve is the pain point of researching for this data. Because when you make decision on clinical trial or product launch, for any medicine, or any medical device that you can never be sure that you got all the information and you're making the right decision, especially the market is moving so fast. And now it's the very first time people have a reliable solution they can rely on.

Michael Fauscette

So, I mean, that sounds great. The data itself, obviously a big part of the problem, because there's so much of it, and it comes from different places. But I know in your platform, you have been using generative AI to help, you know pull all that together and make some sense out of it. So that so that companies can make good decisions. So, can you talk about that a little bit? How does the how does the platform use generative AI to help with that decision process?

Michelle Wu

Sure. That's a great question. So I will break it down into two parts. And the first the why in life science and generative AI is so popular and both the tech world and the industry is so excited about generative AI and second is like how we leverage generative AI to build that. So what differentiate generative AI from every term that we heard before, like AI, deep learning, machine learning, is that for the very first time, generative AI will be able to provide much more richer content than before. Previously for a clinical trial. Traditional AI can answer the question is a patient dead or alive, but generative AI can provide oh the patient much richer content like patient is alive. The vitals increased XYZ blah, blah, blah, and degenerative AI is extremely good at coming up optimize the result with a lot of competing priorities for clinical trial and for any like commercial like partner launch global partner lounge for medicine and devices. There is a lot of competing priorities. Companies want to save the patient but they also want to make money they also want to cut the timeline shorter. So how can you ensure as a manufacturer as a big pharma and medical device companies, how can they ensure that the products they put out in the market are doing the right thing and not hurting the patient. So generative AI is extremely good at solving such problems, like come up with optimized solutions with a lot of competing priorities. So there are as for us how we have been leveraging generative AI is tied to your first question like What do we do and what differentiates us we have built the largest global training set of all the clinical data and regulatory data, data scraping is no joke. And especially when it comes to translation for bowel medical information from Chinese, Japanese and German to English, vice versa. So with that rich data, we're like sitting on Goldmine, we fine tune that our algorithm. So our generative AI is fine tune dedicated to tool life size, like power pharma and medical device. So compare with other generative AI algorithm, they are like a smart student with a bachelor's degree in English literature, they can, if you have a conversation with them, they can have a very interesting conversation with you. But if you ask them to, you know, summarize a medical paper, for example, you gave a medical paper 204 open heart surgery to students in English or literature. Even though they understand every bit of English, they may not understand what this paper means. So that's our fine tune the algorithm. Our fine tune the algorithm is like a smart student, but with MD, PhD, and really understand biology and medicine, and other like industry domain expertise. So what they can do is first day medical writing, every company is required by law and regulation to keep up monitoring the clinical trials, then part of it is done medical writing, one project could potentially take up $1,000,000.06 months to nine months to finish you really truly you need very smart people with MD and PhD and manually read this paper and summarize this paper for you. And what our generative AI can do Nyquist AI can do is to give you a recommendation, whether there's the way very quickly read the medical paper summarize for you and answer questions based on this paper. And we can add that saves. And we can also make recommendations, whether your research you should include this medical paper or not that save at least 50, if not 90% of the time, which is a huge time saver. reduces the burden, like economic burden for manufacturers and hence the patience.

Michael Fauscette

Yeah, I mean, that makes a lot of sense. It's a shift, you know, in decision processes that we've had these expert systems for years that are AI based, but they're very rules based. And it sounds like what you're saying now that shifting in with the generative AI approach is it can actually take that data that you've collected, and project out from that or generate content out from that that actually summarizes gives you a better view of it. And it's not constrained by the rules that actually then interprets whatever the data is that you fed it. That's really interesting. So, can you talk a little bit about maybe some examples of some of the customers like who have used it, you know, maybe in R&D or commercialization or, or even regulatory compliance because it sounds like that's obviously a big part of the industry.

Michelle Wu

yeah. So that's a that's a great question. So, we have seen adoption across the value chain from the preclinical all the way to commercial and post market surveillance, because once that is not like, life science is so unique from other industries. And once you launch the product to say a consumer product on the market, you may not need to require to monitor it much but as long as a part is still on the market, that company are required by law to monitor them. So there's a lot of work going to that aspect as well, for clinical. We have because we have the largest training set, one of the largest ophthalmology public company leverages our data to find, recommend prior art which we hub predicate to help them accelerate the innovation and clinical trial for their global launch previously, so they need to before they run a clinical trial, they need to understand what's the benchmark, and what's the predicate like prior art, what basically, what other people have done in this field. And it's a very niche market is that have to do with children's eyesight, allow children and so it's this part is very near and dear to my heart. I'm very lucky with perfect vision. But I have a lot of friends whose kids are just, they're so adorable, but they have wearing sick glasses. There is medical technologies tool to help with this treatment. So it's very, very interesting and very noble project. Very near and dear to my heart. The challenge is that the company has department to research so many countries and to see what has been done in the past. And it's not, is not known to them that what's the brand was the manufacturers, they have to go from the very fundamental like, what's the medical issue? And what's the basically, what's the indication, what's the technology, and they have hired the three global law firms to do that. And just after six months, you know, like they pay by bureau hours, they're like, Oh, we couldn't find existing solutions for the Asia Pacific market. And then out of serendipity, we got introduced to the VP of Clinical Trial. And he looked into our system, within 20 minutes, he finds the perfect credit cards for China and Japan, other Asia Pacific countries. So that's a huge time saver, let alone like, you know, advertisement for the use case for clinical trials. Go ahead.

Michael Fauscette

That's such a good example of why the industry vertical focus really does add that additional value to the solution. I think, you know, we kind of this year, we spent exploring all the uses of generative AI sort of in a generic way across all the functions. And now, it seems like the real value and you just like you said, I mean, you're accelerating decision making, you're, you're giving them access to data in 20 minutes that they couldn't find in six months. I mean, that's, that's incredible value. And I think that speaks a lot to this idea that next year, that's going to really be the focus for a lot of companies that what what's the vertical focus of a, of a product, that platform that can really deliver value that that makes a ton of sense, I think.

Michelle Wu

Yeah, yeah. I again, like I don't know what I don't know, big nerd and I cause things. My very first job I being in life science, I just constantly realize how different the life science industry is compared with, say, consumer goods or TMT?

Michael Fauscette

Yeah, yeah. I mean, there's, there's vertical use cases, and every one that I could see give, you know, that deliver a lot of value, but in, in an industry that's extremely regulated. And then like you said, they have that downstream responsibility to continue to monitor and invest in ways that a lot of product companies wouldn't necessarily have to do, right. I mean, if it’s broken, they have to fix it. But that's not the same thing as this continuous focus on the product and what the product is doing in the marketplace. So that's, that's a very different use case.

Michelle Wu

yeah, absolutely. Yeah. And across the life cycles of life science paradigm, like medicine drug device, we've seen, like, for example, for competitive landscape previously, again, you need very smart MD and PhD to read through all the papers and decide, oh, is this paradox? Am I a competitor or not, but with generative AI, we can very quickly sift through 10s of PDFs, like the labeling the medical labeling, which is like a dictionary crumbled into a piece of paper and that has been a huge time saver, as well as more regulation which you mentioned, it's everywhere. And the FDA Europe guidance are tremendous value and knowledge and they are more than 3000 guidelines guidance. And they continue to, to roll out and so first the arm No human experts can have knowledge of every regulation and seconds for young people or young generation or new comers coming to this field, it will be so hard for them to decipher those regulations, especially some of the regulations are competing, like contradicting each other. So what degenerative AI can do is like, give you is like a smart AI assistant or AI, in our case, AI scientist who have all the knowledge and help you to decipher what's the what does it mean, what has changed, and any impacts for your business?

Michael Fauscette

And I mean, most of these companies are global. And so I would imagine there's a great deal and deal of variability in the regulations across you know, those different geographies, countries, etc.

Michelle Wu

yeah, absolutely.

Michael Fauscette

So, excuse me, a lot of this relies on the data and the data, you collect the accuracy of the data, that sort of thing. So can we talk a little bit about that process? I mean, how, how do you collect the data? And then how do you ensure the accuracy the quality of the data that you're that you're, you know, that you're using?,

Michelle Wu

Yeah so that's a great question. And data scraping is no joke. So I just gave you a fun anecdote, I thought we have insights into the Chinese market, which is the second largest economy. And given the geopolitical tension, if the IP address is from the US or other countries, it depends on the day, your visits as Chinese FDA websites, sometimes you cannot even refresh the website. So it's a very long process. So we have built first, our top notch engineers around the world. And second, we have a Global Advisory Board, who we do need not only industry expertise, but also countries specific expertise to really understand like, you know, what, what is happening, and because sometimes what you read, the policy could be quite different from the actual implementation. And then we have our we build our proprietary AI clean cleaning and process, we clean the data. And then we have a human expert team to sample test the data to make sure the data is accurate, the data is correct. And then we build our Knowledge Graph to really connect the dots. So dialysis means dialysis, say, adolescents under the age of 14 means that pediatric use and the baby in bed all those synapses and contacts are connected. Last but not least, that when we designed the product, every data point can be traced to the original documents and original contacts. So whenever in doubt, users can always go back to check the original source.

Michael Fauscette

So there's a good bit of human in the loop, then from the quality perspective, at least because of all the different types of expertise you require.

Michelle Wu

Yeah, yes, yes. human involvement expert, the expert. Feedback Loop is very important. This is quite different from open AI as Chad GBT, when they were started out, they outsource it to I heard from to Africa and to leverage the cheap laborers to do the do the human intervention for conversational tool. That's excellent. But for Life Science Solutions, you do need the expertise.

Michael Fauscette

Yeah, that that makes sense. I mean, it's almost like the, you know, the term grounding obviously, means a lot of things in AI, but it's almost that it's almost very specific to the industry, to the fact that you require all of these different experts to be able to interpret and make sure that the quality of the data meets the standards that you're trying to get to.

Michelle Wu

yeah, we also constantly contact as local regulatory bodies, authorities like the US FDA, Europe, EMEA and to get their feedback. And so we are also, like, have a lot of posters. Again, a very nerdy and academic team are very research focused. So we have posters and publications to continue to leverage the community and build a community forward to build the hype hybrid community who are experts in life science, but also understand and embrace technology.

Michael Fauscette

So I know, for example, one of the areas that I've read about and I hear is a very complex is this area around clinical trial design? So I'm curious, you know how using your platform, how does this change that process for your customers and make that more efficient, effective, productive, whatever the right term is there?

Michelle Wu

Yeah so it's, it's almost a universal knowledge that it takes 12 years and 2.7 billion with a, b, to develop a novel therapy. And I guess, given the increasing inflation way to add labor, and everything is going to be even more than just loose change. For pharma and MedTech. The one of the reason that clinical trial is so difficult and so hard is that the tribal knowledge and institutional knowledge go to the person who designed the clinical trials. And what we can provide, again, we have the largest global training set. So if a physician in the US they want to work on, say, Officer Ma, ophthalmology device or some disease, they can very quickly within a few clicks to see all the global summary of clinical trial done in this way, and how do they design their clinical trial, what has worked in the past, and we have all the summaries of, you know, completed trials, and also failure trials, failure trials is so valuable, because you may be able to learn from someone's success, but you can definitely avoid, avoid someone's mistakes. So if clinical trials fail to meet certain endpoint, which means as a result of the clinical trial, that when I, as a new person start to design my clinical trial, I can see, okay, what has been missed in the past, so I can avoid making the same mistake. So that's from a strategic level, from initial research level, and second from Operation level, more than 20 to 30% of clinical trials in the US don't recruit patients at all. And almost 80% of the clinical trials in the US never roll on time. Which means that the 20 to 30% of the facility is just there, they have the medicine, they have the nurse, they have everything set up, there is no patient to enroll this clinical trial. That's, that's pure loss. And, like, the financial sheets for any companies. So our solution also summarizes the performance of facilities and performance of physicians who have prior experience. So there's no guarantee that someone has successfully fully completed oncology clinical trial is going to continue to be successful, but at least the probability of success is, is much higher. So that is that from an operation layer can also is where our solution can also help company to optimize their clinical trials,

Michael Fauscette

I was a few weeks ago had a guest on the show, and we were talking a lot about Data Automation. And, and, and one of the things he brought out and I hadn't really thought of it, and now talking to you, it sort of comes out the same way, in my mind is that, you know, in the past, we had analytics and business intelligence tools, and they're very rules based and, you know, dashboards and that sort of thing, and it's very structured, but it but it leads you only in the direction that you already have planned, right, you don't have the opportunity to be more open and interact with the data. And, and It's more I think that capability to interact with the data in a natural way, is where an awful lot of the value is in, in the system's because it changes analytics into an interactive experience in real time versus that sort of historical perspective. Does that in this context? Does that make sense to you? I mean, it seems like that's a real value inside of what you guys are doing. Is that capability of like, you know, just like you're you said, there's a doctor who's been successful in a certain area is more likely to be successful versus one who doesn't have that experience or had a bad experience. I mean, that's a simple thing, but that probably is really valuable.

Michelle Wu

Yeah, yes. So when it comes to data, trying to understand your question or your comments, The Data Automation lead to analytics that that is no longer robots which make the brainstorming the discovery more interactive. And it's which is also unprecedented in the past. Yeah. Yeah.

Michael Fauscette

That's exactly it. I mean, I think and that's, you know, it's like the idea of how do you transform unstructured data into something that's actionable. And doing it in a way that's natural versus forcing the forcing data into a dashboard that presented only in a specific way, which seems to me like would really limit your capability to do that brainstorming or to do that interaction with the data where the value really might be? But you don't really know that ahead of time, right? Yeah,

Michelle Wu

yeah, that's we have run into situation like this, we were just having conversation with another European, one of the leading European notified body. And because we aggregate all the global adverse event previously, because the adverse events, data is set in such a rudimentary way, you have to like research, okay, I want to see, say, Pfizer, Medtronic, and other companies like adversity, but of this particular product. So you have to know you already have a bias built in, when you go down the path of exploration, to like, you have to know what you are looking for, in order to look for what if they, you have all the data and generative AI, they can summarize it for you in a way that you haven't been experienced before. So when we presented our solution to this notified body, they were like, wow, you just finished your you know, we have been doing this for 20 years, in the old way is like, Michael, you're on the East Coast, like previous will, people are used to drive their cars, all the way from East Coast to pull out to Silicon Valley. Now they can just, you know, get on like, like a driverless car by Tesla, or even just like get down. Driverless playing and transport to Silicon Valley is just a totally different way of looking at the data. And we also, right now the technology is so ahead of not only not only with the use case, but also with the adaption curve of a tech adoption palette for the industry. I think that at least from my experience, we we constantly have five to nine inbound calls from one person startup, all tech startup to like multinational corporations, they are constantly got to the emotional or psychological shock, that the things that they used to do like the dashboard has been existing for 20 years or 30 years. And now they just need to have a conversation with the bots. And the bots can write them a beautiful summary report and even like PowerPoint, right,

Michael Fauscette

yeah, so it puts everything together and then displays it in a way that's digestible to you

Michelle Wu

at a time and speeds that no human bodies are capable.,

Michael Fauscette

Well that kind of relates back to something you were talking about earlier that I am a little curious about, because I also hadn't thought about this in this context. That's this idea of competitive analysis. In, you know, applying generative AI for competitive analysis, especially in a field that's so complex and technical, as you know, as life sciences, taking that ability to summarize and help make sense of data, and then funding those specifics of what is a competitor versus what isn't, and how that could actually inform your decision process as you're thinking about how you structure and what direction you go with your product.

Michelle Wu

Yeah, it's very different. I've seen that actually, I think I've seen bigger company taking advantage of this opportunity. And some of them are even trying to train their own charity that I model based on huge volume of data. If you think of this way, if you look at industry this way, Michael, like the Pfizer's and Medtronic’s of the world, they have, like Merck has been around for 130 years, that's 100 years, three years was of data, clinical trial patient information. Any big pharma company and medical device company could be a data company and AI company in the future.

Michael Fauscette

I mean that, again, that that volume of data from all of that historical information and the ability then to take that and make some sense out of it, that's, that's actually really, a must be really exciting and obvious when you talk to companies like that, how they could actually apply this in a in a very different way. So I'm curious about the future then. So I, I mean, I see the depth of vertical focus. And I think that's really exciting. But where does this go? I mean, what do you think, over, you know, some period of time into the future? Let's project a little bit, what do you think the impact of generative AI is and how it evolves in in life sciences over the next 510 years?

Michelle Wu

Yeah, that's, uh, ah, so I think Bill Gates once that's people tend to overestimate what you can achieve in like five days versus what you can achieve in five years. So first, generative AI for life size is no longer just a buzzword, people are proactively adopting those technologies and tools and software's either in their sorry, apps in their real life. And second, I do see a change in the form of organization. It ties to your previous question, why clinical trial is so complicated, why like drug discovery is so expensive. Part of it is that they are just sheer volume of information. For example, dropout from Stanford, a 17 year old 18 year old prodigy could code just as good as some, like PhD see in CS. But a doctor with 20 years of clinical practice is fundamentally different from a fresh graduate from medical school. Because the industry requires so much institutional knowledge and expertise that to some degree slows down the innovation curve. With generative AI, what it really provides for the industry is a speed and precedent speed to digest and make sense of huge volume of unstructured data that human brains previously cannot achieve. I think it will, it will change how the organization looks maybe in the future. Customer Success or like customer support for drugs or medical device will be upgraded to a chatbot. And then there will be new jobs created new roles created by generative AI in life science. Yeah.

Michael Fauscette

I mean, and we see that in other industries, too, the more you can build out the end customer supports a great example where you have an assistant that that can either solve problems, which is great, right? So you can in some interactions with customers in a positive way very quickly, because they have access to all the data. And then they also have the capability to make your customer service agent able to deal with more complicated issues that do need human interaction, but they can still provide them that data access. That's, that's really interesting in it. And it makes a lot of sense in this context, too. Because of the complexity. Yeah, yeah. Yeah. Yeah. Well, so we're running out of time, although this really interesting. I could, I could keep going. But I guess at some point, we probably have to give the audience a break. So I the one thing I always ask at the end of the interview, though, and I would love to hear your recommendation on you know, someone that the audience could, you know, research listen to read a thought leader or somebody in this field that you think is really, you know, educational and could really help people understand more about generative AI and how it's being applied to business.

Michelle Wu

Yeah, yeah, I actually have two recommendations. Now you prime my brain. One is my Stanford professor Andrew Ng and what differentiates Hear from all the other top researchers in AI field is that he's so far the only researcher that I've known that can convert very complex, very technical concept into layman's terms and use very vivid stories that you can relate to the technical change. And the second one is Kai-Fu Lee. And he came from Google, and he's become a venture capitalist and become an entrepreneur, what he can, what is unique about him, he is like, he can tell a fascinating AI story in almost a science fiction context, which is very intriguing, at the same time, deliver behind the facade of like science fiction stories, which is very intriguing and vivid. It delivers impact value for the business.

Michael Fauscette

I mean, that those are great recommendations and Andrew Ng I read a good bit of his stuff. And I think he's, you know, extremely interesting. And the idea that I think this is something that's really important today, the idea of taking this technology and translating it into business terms so that people can understand consume, you know, actually be able to apply it. So that's, that's great. Thank you. And Michelle, thanks so much for joining today. Really interesting conversation, and I really appreciate it.

Michelle Wu

Yes, thank you, Michael, for having me.

Michael Fauscette

And that's the show for this week. Thank you all for joining remember to hit that subscribe button. And for more on AI and other software, research reports and post check out the area on research.com/blog and slash research reports. And don't forget to join us next week. I'm Michael Fauscette. And this is the disambiguation podcast.