How AI helps women track hormonal health, with Marina Pavlovic Rivas

0:00Starting a business comes with its share of ups and downs, which is why staying true to your vision is essential. a non-negotiable for Romeo and Milka Bregali, Capital One business customers and co-owners of Ra's plant-based restaurant in New York. Romeo and Milka took a leap of faith when starting their own restaurant, gutting an empty space and building it from the ground up. Every pipe, every wall, every detail. But building from scratch came with a heavy financial burden, which is when they turned to their Capital One business card. With the flexibility of the card's no preset spending limit, they were able to spend more and earn more rewards while bringing their vision to life.

0:36Today, Raz's success is proof that with passion and the right support, it's possible to make your dreams a reality. Learn more at CapitalOne.com slash business cards. I was experiencing different hormonal symptoms, that there's different symptoms that can happen throughout the month, throughout the year, from year to year. When it comes to hormonal contraception, getting off hormonal contraception throughout the menstrual cycle, and speaking with hundreds of women around me, realized that it was very common. Marina Pavlovak-Rivas wanted answers to her questions around hormonal health. And to do that, she needed data.

1:22As a data scientist, Marina realized how little she actually knew about her own body. But the problem didn't stop with her. There are huge gaps in research when it comes to women's hormonal health. So very often when we think of hormones, we think of fertility. And for sure, that's part of the picture. But it goes much beyond that. Hormones play a role in all bodily function. But until now, they didn't have the information to be able to see how those fluctuations ultimately impact health. Marina wants to change the way we think about hormonal health. She's co-founder of Ellie Health, a company making the first at-home monitor that tracks your hormonal levels using saliva.

2:09Women's health, and in particular, hormonal health, is something near and dear to my heart. It affects so many of us, yet it's a conversation we aren't having. Today, we're changing that. Marina and I are talking about the roots of the women's health gap, how novel sensors and AI can democratize access to health, and why we need this longitudinal data. I'm Rana El-Khalyubi, and this is Pioneers of AI, a podcast taking you behind the scenes of the AI revolution.

2:49Hi, Marina. Hi, Rana. Welcome to Pioneers of AI. It's so great to have you here. Thanks for having me. So today we're going to talk about women's hormonal health, which is, of course, one of the most powerful yet most misunderstood aspects of women's physical, emotional and mental well-being. And it's something that impacts our fertility to our mood, our longevity. It's so underfunded both in research and in innovation, which is why I'm so excited to have you on the show and also learn about your company, Eli Health. But before we do that, I would love to hear more about your story because I think you have a quite an untraditional route into tech.

3:30Yes. So my background is in data science originally. Prior to that, I was in communication. That combination may seem unusual, but for me, it was the same passion of putting information in the hands of people and making it available at scale. I had the first company in the field of AI after completing my studies. And after the acquisition of this company, I started LE from a personal need. Of course, with a background in data science, I value having data when it comes to making important decisions. But I realized that when it came to my health, hormonal symptoms, that this data was missing. So this is what ultimately led to the creation of the company with my co-founder and making this tool accessible for the first time.

4:26You know, women's health is a topic that I'm personally very passionate about, and I'm committed to highlighting amazing founders who are building in this space. And we actually had Alicia Chang-Rodrigue, the founder, of course you know her, the founder of BloomerTech. And she talked about her smart bras for cardiovascular health for women. And we spend a lot of time talking about how is it, how come, you know, women's health is so underfunded and so misunderstood. So why do you think this is the case? There's different reasons behind that. There's more systemic reasons, of course, when it comes to different disparities, when it comes to gender.

5:09until not so long ago, women were not included in clinical trials. But there's a piece of that that is also around, historically, the perceived complexity of women, that it was considered too complex and expensive to include women in clinical trials, given those fluctuations and hard to isolate variables. So that's a piece of it. But more broadly, when it comes, for example, to startups and funding in the space, the presence of women, both on the founder side, but also on the investor side, plays a big role. Because when it comes to women's health, of course, having investors that understand firsthand what those challenges are makes a big difference in being able to seeing an opportunity, especially when it comes to creating a technology that never existed before.

6:12Having that firsthand challenge enables investors to see that future much more easily. I always say who writes the check determines who gets the check, right? And if you're pitching to an audience who don't really get the pain point and they think it's a fringe problem, right? But of course, like women make up like 50 % of the world's population, so it's not really a fringe problem. But if they haven't experienced it, they won't really see the pain point. And we see often as founders, many of us in the space receive the comment that women's health is a niche. When it comes to healthcare buying decisions, 80 % of the spending is controlled by women.

6:54So it's far from being a niche. So from a venture standpoint, that underrepresentation also represents an opportunity for investors out there that understand that gap. So we kind of already talked about how, for women, our hormones influence every aspect of our well-being. Yet for most of us, and I'll speak for myself here, you know, you go through life without actually any understanding of your hormonal patterns and the effect it has on your health. And at best, maybe you do a blood test once or twice a year. It gives you a snapshot of your hormonal health. But it's not really actionable because, of course, our hormones fluctuate on even a daily basis.

7:39And so you're on a mission to change that. So tell us your approach. Exactly. So that was part of the insights that led to the creation of the company. Similarly to how it wouldn't make sense to measure your heart rate once a year and then make decisions all the other days of the year based on that information. It's the same for hormones. Having that single snapshot can, yes, tell you information. Sometimes it's enough. A first data point that can lead to a diagnosis, for example. But when it comes to using that information to optimize your health and wellness in a way that you can see how this information can guide lifestyle decision, that single data point per year then becomes insufficient.

8:29Similarly to how measuring your sleep once a year, would it make sense to see how your actions on a day-to-day basis impact your sleep? Same for heart rate and other bar markers. So we're introducing the same for hormones, that possibility of having the data frequently over time in order to see how your different lifestyle decisions impact this data. So you kind of honed into measuring cortisol. I'm curious why, and maybe you can use this as an opportunity to talk about some of the other hormones that could be of interest as well. Yes, so currently we're measuring cortisol and progesterone. And we started with those two hormones because of the key roles they're having for health.

9:19Cortisol plays a big role across all bodily function. So, for example, if cortisol is unbalanced, it becomes very hard to balance other areas of health. To give a couple examples, when it's 20 % higher than baseline, that can be enough for ovulation not to happen. And as a result, not being able to conceive or having different challenges when it comes to sleep, mood and other areas of health, It can lead to metabolic imbalances that then lead to weight gain and same for cognitive health, physical performance. So starting with that hormone plays a big role for those other areas. And often we refer to cortisol as the stress hormone, but more and more people call it the master hormone because of that role it has really orchestrating the different systems of the body.

10:18And for progesterone, it's one of the key female hormones. So being able to have this information throughout the month, throughout the years, then enables women to have that visibility on this bar markers that plays a big role. So let's talk about the form factor you landed on and why. Yes, so we used saliva. And for us, when we started the company, we had two, more than two, but two key guiding principles was to have a test that is as easy as brushing your teeth. And that can be as affordable as a cup of coffee. And the reason behind that is to enable that product and that technology to be a key piece of a daily routine.

11:06So we've explored different form factors going from patches to contact lenses. And we went really broad into the different possibilities and landed on saliva due to its non-invasive format. And also the possibility to do that testing anytime, anywhere. We have users that are doing it in their car, on their way to work, at the gym. So when it comes to seeing how this data correlates with lifestyle, that possibility of having information instantly and wherever you are plays a big role in choosing that form factor. In a minute, Marina walks us through how this test actually works. Plus, we talk about how computer vision can be an integral part of personalized medicine.

12:03Stay with us.

12:15If you've spent any time building AI products or leading technical teams, you know this. Transformation doesn't fail because of ideas. it fails because teams can't move together. Enter Atlassian's teamwork collection. It has planning in Jira, documentation in Confluence, video updates in Loom, and now AI agents in Rovo, which connects the dots across your work so nothing gets lost. It's one AI-powered teamwork platform designed for how modern teams actually build. Learn more at Atlassian.com slash TeamChanger. That's A-T-L-A-S-S-I-A-N dot com slash teamchanger

13:05So walk us through how this would work So say I order a Lee Health kit and it arrives at home What do I do? Yes, so can I show one to you? Yeah, great At this point in our conversation Marina took out one of the LE tests. It's about the size of a pregnancy test you'd pick up at the drugstore. It kind of looks like one too. So you receive it in a pouch like this one. This is the test. And you put this extremity on your tongue for a few seconds. Pull on the strip and the saliva starts flowing. And when it's fully developed, you see lines appearing on the test. And you take a picture with your phone and that picture is being translated into the hormonal levels, but also what they mean for you in the context of your own baseline and your own goals as well.

14:07So the app's going to do the analysis and we're going to dive into the algorithm behind that in a second. But then how is the outcome represented? What does it say? So on the app, you see similarly again to a continuous glucose monitor or a smartwatch, you see your hormonal levels, but also compare it against benchmarks. So you can see whether it's within range or outside of range against the population, but also how it fluctuates versus your own baseline. because you could be within the normal range, but still fluctuate from a day to another. So we highlight those variability and also provide different metrics like score and recommendations around what you can do based on this pattern.

14:57So then you can create that feedback loop to see the actions that you implement. Are they actually affecting your cortisol or progesterone curve? Okay, I have a whole bunch of questions for you. So I'm a fanatic, like where I wear my whoop, like religiously, because I love the data, and it's really insightful and actionable. So can you kind of share some of the recommendations that Elie could give one based on the tests? Yes. So when it comes to the recommendations, we focus on lifestyle. And those lifestyle categories are the classical ones that we've all heard about when it comes to exercise, sleep, food, other actions like light exposure, sleep consistency.

15:47However, the devil is in the detail. For example, when it comes to exercise, it's not simply about doing more exercise, let's say. If you have a cortisol pattern that is, let's say, too high in the evening, then it would be not recommended to do high-intensity exercise later in the day. versus if your cortisol is too low in the morning, then it can be recommended to do a higher intensity exercise early in the day. That's just two examples of how in the same category, the type of recommendation when it comes to exercise intensity, timing, recovery, changes based on the patterns that are being observed.

16:34You mentioned a continuous glucose monitor. And of course, the hormonal tracker equivalent of it would sample your hormone levels multiple times, say a minute. How often should one take a saliva test so that you're getting an accurate enough picture? We recommend at least four times throughout the month. And that's really a minimum to see those fluctuations over time. So for cortisol, we're interested in seeing the levels in the morning and the evening. So one test in the morning, once in the evening. And to do that at least today throughout the month. And for progesterone, given that the fluctuation happens throughout the month, those tests are across the month instead of being within the same day.

17:23And that's a baseline. But what we're seeing is that depending on the lifestyle and depending on the goals, there's people that want to test much more frequently than that and others for which four times is enough. You know, also similar to the continuous glucose monitors, obviously it started first with diabetic patients and now it's becoming more popular just for the average consumer, right? Who is your target customer for this kind of test? Yeah, so what we saw in the industry when it comes, for example, to glucose, it first started in clinical settings to then evolve towards consumers. And then there's other technologies like smartwatches that started with consumers and that are now evolving towards clinical applications.

18:14We are more following the second path, starting with consumers and eventually evolving towards clinical use case for the simple reason that people want to have access to this data and have access to the information directly in their hands. So similar, I would say, type of behaviors than people who love seeing their data on their smartwatch, but that also understand that this is not enough for the full picture. Do you focus on a particular segment, like women who are trying to get pregnant or perimenopausal or menopausal? So we're seeing a lot of peaks around perimenopause, perimenopause and overall health.

19:02Because during that transition that can last more than 10 years, there's a lot of symptoms. There's a lot of changes that are happening for the first time. And we see a big demand in having access to the relevant information during that phase in order to be able to manage symptoms and act proactively around those different areas. So we're an AI show. So I want to now take us behind the scenes and talk about kind of the R &D that was involved in building the product and the role AI plays in all of this. I've already heard one area where I believe there's going to be an AI role, which is the computer vision algorithm that takes that strip and translates it to a score.

19:48So tell us more. So from an R &D standpoint, this was a very multidisciplinary effort. Of course, the chemistry itself was a big portion of that, the microfluidic aspect, the hardware. But when it comes to AI, as you mentioned, the computer vision detection was a big portion of this. Initially, we used to have a dedicated reader into which we had to insert the test to take the reading. So this had, of course, a couple downsides, including the price for the user and reducing ultimately the accessibility. So replacing that reader with the smartphone camera was a big step in the direction towards making this product accessible at scale.

20:37And that's something we couldn't have done without computer vision algorithms. So that's really a key role when it comes to AI for our system. And then, of course, the data interpretation, because those computer vision algorithms enables us to get the result, which is great. But then the next step in that process is to provide the interpretation of that result and also the different insights and recommendations that can be a next step to that result. So I'm a computer vision scientist by background. And so I can see some of the challenges that may come up. For example, if I'm doing this test and it's low lighting, right?

21:22That's going to affect the image the camera is going to take and may affect the results. So how do you guard against some of these, like, I don't know, lighting conditions, angle conditions, blurring photos, all of these challenges? Well, it was, as you mentioned, the key elements that made this a challenging thing to solve. We were unsure when we started that development if we would even one day get to a similar performance to the reader. And today we have the same due to the different datasets that we've collected in order to train the algorithms in a robust way. So that came, for example, using different cameras, using different lighting conditions, different angles across the different concentrations that can be found in saliva.

22:15So that's one thing from the training standpoint. But then from the user interface standpoint, there's also different guardrails that enables to guide people. For example, how the picture of the test is being taken if it's not aligned properly. There's a message that appears on the interface to guide users into placing it the right way. Same for lighting. If it's too dark, there's information that guides users into what those minimal settings should be to take a picture. If any of our listeners, you know, do the mobile check function in your mobile bank app, I think it's very similar where you're trying to take a picture of a check, but then it'll guide you to position it right and ensure that the lighting is correct.

23:10What's next on the roadmap? Like, are there any other hormones that you're particularly interested in having them be tested or included in the product? Yes, there's many other hormones that we plan to add in the pipeline, including eshodol and testosterone. Testosterone, which often when we think of that hormone, we think of men. And of course, it's a key hormone for them. but it's also playing a big role for women's health which is something that is not always known so that's one we're particularly excited for the impact it can have for women's health but also eventually for men and as we we saw with our early users you've asked earlier one of the surprises that we got with that early usage and seeing that men also have a big interest around testing their hormones.

24:04That was something we were excited to see. That is so fascinating too, right? Because why not? Exactly. We're going to take a short break. When we come back, the Healthspan revolution and why we need better data sets.

24:37Meet Nicole Nicholas, Capital One business customer and co-owner of Ansett Uncles, a plant-based restaurant and community space in Brooklyn, New York, that got its start from a need for unity. The inspiration, it was born from the desire to create a space that felt like home, where we can connect community culture, good food, and come together with family and friends. That's how we birthed aunts and uncles. Nicole and her husband, Mike, were fulfilling their dream of bringing people together out of their home kitchen. But they soon learned that the demand for community was greater than they knew.

25:08It became overwhelming and we were like, we need home, but not in our actual home. We realized that there was also a need in our community for something bigger in our neighborhood. So we had to find a place. Moving from a home operation into a storefront was a huge next step. But Nicole and Mike were able to take it on with the help of Capital One Business. It's not for the weak. As a small business, finding resources is super important because that's the way you'll be able to manage and scale. We would have never done that without having Capital One to be able to help us along the way. The cashback rewards are very helpful.

25:44You know, it just gave us that runway to be able to breathe a little bit. Then you get to focus on the cooking of the food and making the experience great. To learn more, go to CapitalOne.com slash business cards.

26:01So one of the reasons I'm super excited, you know, to have this conversation is I really do believe that we are at the cusp of a health span revolution. And it's powered by this trifecta of sensors in general becoming much more available. And this multimodal data explosion, right, and longitudinal data, which you're collecting, and then combining that with both predictive and generative AI. So I am curious, what are you seeing from your point of view and what gets you excited about this space and what other innovations are you seeing? Yeah, so some of the things we find extremely exciting is, as you mentioned, the combination of that increasing interest around preventative health, longevity, but also combined to increase access to the data.

26:50And that was one of the reasons why we started this company, because our hypothesis was that to be truly personalized and truly preventative, you need to have a frequency of data that matches the frequency of life. And without this granularity in terms of information, our belief is that the algorithms can be, of course, only as good as the data that they're being fed with. So if that data is missing, then there's some major limitation when it comes to that vision of having health in real time throughout our life. So some of the things we're extremely excited about is seeing more companies creating this bridge between the biological and the digital, especially in an era where models are more accessible than they ever been.

27:47AI is now being accessible to the mainstream, which is amazing. But the data is still missing around some areas of health. So we're seeing, as counterintuitive that it may sound, that the heart tech components or really physical products that make that bridge will be a key driver in that health revolution. Now, you're also building, I believe, the history's first large-scale hormonal data set of this kind. So two questions there. How are you ensuring that this data is protected? Because it's, you know, very personal data. and two, I'm just curious, what are some of the questions that you would love to dive into once you've got that data at scale?

28:32Yes, well, first, when it comes to privacy, we put the control in the hands of users where they have the possibility to not share some data and not log the data they don't want to log, but also ultimately in the background, making sure that data is structured properly, which means in an anonymized way. so that when we do those analysis, it's never linked to a specific user. From a data set standpoint, what we're really excited about is how it will help uncover many of the gaps that still exist today when it comes to hormones. So seeing, for example, how different interventions can be powerful to change how someone feels on a daily basis.

29:25But beyond that, one of the areas that we're extremely excited is this ability using data sets around hormones to prevent different conditions. There's already some early data that shows, for example, based on the patterns observed around certain hormones, that it can be predicted to an increased risk of bone-related conditions like osteoporosis, heart condition, cognitive conditions. And so being able ultimately to see how a certain pattern can lead potentially to those different outcomes and act preventatively to change that trajectory is something we're extremely excited. Yeah, amazing. All right, last question.

30:15And it's a question I ask all my guests on the show. And it's one I think a lot about. So as AI becomes more powerful, you know, smarter, more creative, even more empathetic, which is what I spent the last 25 years building, what do you think makes us human in this age of AI? That's a great question. I think ultimately there's something special around human creativity. For AI, it's different than other tools that we've saw in the past. sometimes there's that analogy that we've been through similar transformation in the past going from let's say a typewriter to a computer to now AI. I do find that fundamentally it's different that AI is not just the brush it's also a piece of the painter but the intention behind the process stays human.