Driving health breakthroughs with integrated data strategies & gen AI

Hello, everyone. Um I will say I'm not as caffeinated as I should be. So, cheers applause, excitement.

My name is Valerie Delva. I lead data and ML strategy and solutions for healthcare and life sciences.

Um I'm really excited that you guys are joining us today. I'm gonna breeze through some things, stop in some areas to talk a little bit more slowly because what I'm hoping to discuss is how do you build an integrated, more than just an end to end an integrated data strategy that helps you unlock ML and generative AI use cases and really think about how do you not just innovate but scale that innovation in a sustainable way.

I hope that feedback was hit my mic.

So what will we cover? Um it's detailed here on the slide, but a theme that you'll hear is the importance of data as a critical differentiator to accelerate those insights in that innovation. So first I'm going to start with the chat challenges and the opportunities that we're seeing in this space. And then I'm gonna talk about our AWS health free data portfolio, the products, the solutions that we have some of our partner offerings and then specifically what does an integrated strategy look like and how do you put that into action?

But before we start there, i often like to start with this slide, taking a step back and asking each of you to keep top of mind the why behind all of the work that we do, especially if you're in health care, how can we together scale and accelerate the future of health? And this is one where it's genomics, ML and AI, driving more addictive preventative, personalized and participatory medicine for all.

So I very much believe that it's not in two decades or even a decade from today. But right now, we can really think about the power of the emerging technologies and the data that we have in place to get to this uh future vision for folks.

So over the past several years, I will say that there's a clustering of four or so trends, uh challenges or opportunities, depending on where you sit, you'll call it a cha a challenge or an opportunity for health care organizations in terms of how you create store and use data.

Um and until quite recently, you can make the argument that you could not easily address many of the ones you see here on the screen with the technologies available. And I'm gonna go through each of them one by one and probably spend a bit more time on the fourth one. Gen AI as you guys can imagine why.

Um but to start many of you are asked to speed up how you discover and development and develop new therapies. For example, if you're sitting in the clinical research lab in an academic medical center or working with bio pharma companies on developing the new therapy, it takes almost a decade and nearly $3 billion to bring a new drug to market.

So imagine the tremendous patient impact this has but then also it's economically unsustainable. So all of this means that you need to make better use of the data that you have to speed up early research, clinical development, clinical trials, but also prove efficacy uh to payers other bodies including through the use of real world data.

So um challenge number one is how do you find and use data faster while at the same time reducing costs?

Second. And I think we're all aware of this. The volume and complexity of data are both exploding. So for example, your labs can now generate multiple terabytes of data from newer assays like cell painting RNA sequencing or even cryoEM yet.

And this is a stat that i know a lot of us have heard before, but i like repeating it to make the 0.97% of data generated today is unused because it's not in a manner that makes it easy to find or to use. And i'm going to come back to that some solutions, we have to address that. But that is a staggering statistic. If you think about it on top of that, scaling up the storage for these data types would be hard enough if they were all the same type of data. But as you know, they're not and they are often different data modalities as you see on the right.

So now let's continue on, we've talked about different data modalities, the volume and complexity of the data and the time it takes to develop new medicines. We're also realizing that compounding on top of all of that, every year, there's a growing percentage of new therapies that reach the market that are originated, not just from one company coming up with it, developing a pipeline, but a collaboration between a bio pharma company and a third party such as um a start up a university lab, an academic medical center, uh population research partner. Uh more than 50% of drug launches today now originate in some form of cola those type of collaborations.

So if you think about it challenge number three is how do you more efficiently discover what data your collaborators have that you're allowed to access and then make appropriate use of that data in a secure way.

So this all means you need powerful tools for data cataloging data governance and data sharing. And i'm gonna spend some time on that. But if you remember the fourth challenge and this is what i'm going to spend a couple of minutes on generative AI and to put all of us on the same page when i say generative AI, i'm speaking about a type of artificial intelligence that can create new contents uh ideas uh whether that's conversations, stories, images, videos, music, protein structures, right?

So the promise of generative AI in health care is vast. But in addition to that, um the potential to accelerate innovations and increase efficiencies across the value chain is also large. But what we realize is that to efficiently manage large volumes of data needed for such models. Uh for generative AI, they are important considerations, especially if you're in health care that you need to keep top of mind.

So for example, if you're working with patient data, proprietary data, you have to think about not just keeping it private and confidential, but ensuring the right people have the right level of access and controls at the right time, but also thinking of data sovereignty and data providence um considerations.

So AI only intensify some of these trends and opportunities that i've been calling out. And what do i mean by intensify? What i will call out here is that data is the difference between generic generative AI applications and those that truly know your business, your patient and your workforce deeply.

So every company at some point will get to the place where you have access to the same foundation models. Let's assume that's a true statement. But companies that will be successful in the long term in building generative AI applications with real business value and impact for patients are those who are doing so using their own valuable data.

And these companies that have not yet found ways to efficiently harmonize and provide ready access to their data will be unable to, for example, fine tune generative AI to unlock its transformative potential.

So especially in health care, this is so relevant where the data and use cases are so specialized and the amount and types of data are growing exponentially. If we just think of genome data alone, it's expected to reach 40 exabytes in the next decade.

The other point that i want to highlight here before i move on to well, what is this integrated strategy look like generative AI does not promise instant insights. The key to delivering health care use cases with impact is quality data. And i stress that because not just because this is a conversation on the impact of data for ML and gen AI. But the fact is the number one challenge for many organizations in health care is realizing the potential of generative AI data systems are sprawling uh siloed complex uh diverse data sets uh spread across data, different data lakes, data warehouses, different databases, SAS applications devices and on prem systems.

So to get meaningful insights, you need a clear data and infrastructure strategy. And it's all to say that quality data matters. But at the end of the day, your data is your differentiator.

So then the question i often get is ok, you've highlighted all of that. But how do, what are the steps into an integrated data strategy and what does good look like?

So the first thing i'll call out is an integrated strate data strategy does not just mean finding ways to break down the silos that exist even internally. Yes, i would say that's step one, a step one b would be. How do you ensure that you're doing so in a manner where there is secure government access that's put in place. So only those who need access to the data have access to it and importantly, not just breaking down the silos with the right governance, but going back to the value delivered in time to value.

So really thinking about identifying which use cases are critical to driving business outcomes for your organization. And on top of that, the types of data needed to execute on those use cases. I find if you're not thinking about breaking down the silos, the right governance and the right use cases, you don't have the start to an integrated data strategy that will be sustainable and successful because all data are not created equally.

So for example, perhaps you're a pharma company, collaborating with clinical research on clinical research, with an academic medical center or a hospital uh in some manner, you need to augment your own data with insights hidden in patient medical records or research data sets. For example.

So your data strategy would need to include seamless integration with real world data assets in a secure privacy preserving environment purpose built for third party collaborations of this kind.

So in addition to a comprehensive range of analytics, ML and generative AI capabilities, your goal is thinking about how do i go not just from having information and data but insights and evidence generation.

So ho how can we help? What do we have to help address? I set up? These are the challenges and opportunities. Uh we realized several years ago that continuing to use general purpose uh cloud services alone would not get the job done. And importantly storing, analyzing cataloging and sharing multimodal health data requires purpose built services and solutions, not just from AWS but also our partners.

Uh so i'm gonna cover a few of these. Now. Um the first one i'm going to start off with is storing and analyzing multimodal data. And on the left, you'll see health link, health link enables you to store and analyze text based clinical data such as electronic medical records, uh claims data and a key feature. It automatically converts unstructured clinical data like physician notes into structured data and it provides a chronological view of individual or patient population health data.

Additionally, our um health imaging service that you'll see here. Next can reduce medical imaging storage costs by up to 40%. Uh what i think is really interesting about it is it can allow multiple users to access the same copy of that image at the same time with sub second retrieval.

Now, the third pillar which you may be familiar with is healthomics. You can transform genomic transcript genomic transcript omic and other omics data to scale, generate uh and generate insights, improve health and advanced scientific discoveries.

What i think is the most important thing as we think of these services and some of the other ones i'm going to cover shortly is all of these can make their data available for analytics and machine learning in a matter of minutes and not just as separate data types but in a multimodal manner.

Finally, one of the newer services that we announced and it's HIAA eligible as of a couple of weeks ago is house, it empowers health care software vendors to automatically generate transcripts, summarize notes and clinical insights by analyzing patient clinician conversations in their clinical applications.

Now, let's talk about data collaborations and cataloging linking and accessing data across all teams in research and clinical development. For example, users often start with three questions and these are the three questions that my teams get a lot.

Um who has the data that i need? Am i allowed to access it? And if all those two statements are true, is it in a format or stored in a manner that makes it useful.

So to start on the left, um i'm excited to share that. Just last week, we announced AWS Entity Resolution as a HIPAA eligible service. So it allows you to easily match link and enhance related records across multiple applications, multiple systems, multiple data stores using flexible and configurable workflows that take only minutes to set up.

Uh i also hear from the same groups in research and development and clinical organizations. They need to access real world data, whether that's to study biomarkers or disease pathways or to design a clinical trial.

So many of our customers spent hundreds of millions of dollars and hours and weeks and months trying to procure real world data. Our goal is to simplify that process um and streamline it to minutes if we can. So that's why we created ADX now while ADX is helping more and more customers every day, researchers and health care organizations sometimes do not want to or cannot share their data or put it onto an exchange.

So that's why we announce cleaner rooms to, to address exactly these needs with cleaner rooms, you can securely analyze, collaborate and without sharing or revealing your underlying data. And if you heard Swami's keynote yesterday, you probably also heard a new ML feature of clean rooms that allows you to do look alike modeling.

Besides clean rooms, many of my conversations also go well, accessing and collaborating on data externally is important. But a huge pain point for me is how do i even unlock data within my own organization? How do i know what i have?

Um as most of you already know, finding out the data that your organization already has and allowing appropriate users to access that data can be incredibly hard. Um and the reality is there are several great catalog offerings out there today and including on AWS. But back to the point that i made earlier, health care use cases are not sufficiently addressed with general purpose services. And that's part of the reason why we created um datazone that you see here now.

So customers are leveraging amazon datazone to build an enterprise data ecosystem like a data mesh platform with federated governance where multiple personas across domains like r and d and commercial can collaborate at the project level.

Now, we have an entire slate and suite of services also purpose built for generative AI and ML. What i will call out here on the left and spend some time on is amazon bedrock. It's a fully managed service that offers a choice of high performing foundation models from leading AI companies.

So the ad a is we want to give you choice AI 21 labs, Anthropic Cohere Meta Stability AI and also uh models of our own coming from Amazon along with a broad set of capabilities that you need to build generative AI applications, simplifying development while also maintaining privacy and security.

The other thing i'll call out before i go to my next slide is our titan service is comprised of powerful general purpose models to build a support, uh to build and support a variety of use cases. Here is the remaining formatted transcript:

I've covered a lot very quickly and often i get the question of how does it all come together? You've given me the elements of an integrated data strategy. You said, what good looks like, what the services and what the options are? And i like to call out this slide and i call this out because an integrated data strategy means going from data to actual insights. Is it clinically useful and meaningful?

Um it means thinking about how do you make it easier to find access, collaborate and perform multimodal analysis but not just to have data to have data. Sit there. I wanna call out the right side of the slide. The use case matters, the value delivered matters.

So whether it's to detect diseases earlier to design better clinical trials, development, develop medicines faster or at the very top, it's not just about predicting and analyzing but is there a way for us to track disease progression at the individual or population health levels to think differently about health and care management?

And i will go back to uh i guess where i started this on the future of health. Together we can think about these challenges and opportunities and how do we turn them into impactful solutions, not just individual services and offerings to drive better outcomes for patients.

Think about all the work that you guys are doing now. Um whether that's in diagnostics or targeted drug development and delivery, how do we get to a place where we're thinking of rural evidence generation and uh changing the definition of health and wellness management.

I think we're on track to make real progress towards precision health. And if your next question is, well, how do i get started? Where do i go?

Um i'd say pay attention to the screen here. Um if you need peer level executive sparring partner, the first box here on the left, um we have a team of data strategists that can help you think about and they're comprised of former CXOs.

Um whether you're a large enterprise, a start up wherever, whatever your size is um to help you think through that. So then if you want to build out a data strategy and a vision or if you're like, well, i'm not even there. I'm at the point of thinking about how do i modernize and migrate uh my infrastructure.

Um so there are a slew of options um in addition to our AWS AI innovation center that's here and ready to help it.

I put this slide up as a reminder of not just the work that we do every day, but also the impact that we can make across a few different areas. And in this case, in particular, i want to call out that we've partnered with the children's brain tumor network uh to showcase artwork by Cameron. Age five.

Cameron chose to draw us all a picture of her swimming to showcase your super power. Um you can learn more about the children's tumor network and advancing pediatric cancer research and get your own pin if you visit our lounge that i believe is nearby.

Um and there's an entire slew of demos uh given all the services and solutions that i've talked through that you can visit as well.

So with that, that's my contact information. I will take any questions. Um and thank you for the time.