Episode #239: Chris Fernandez, EnsoData, “There Was One Particular Area That Felt Like It Was At Least 5 Years Behind…And That Was The Application Of AI To Waveform Data”
Guest: Chris Fernandez is co-founder and CEO of EnsoData, a startup with technology using artificial intelligence and machine learning to analyze wave-form data to save clinicians time on labor-intensive, complex data interpretation and help them across the care continuum.
Date Recorded: 7/1/2020 | Run-Time: 50:12
Summary: In today’s episode, we’re talking all things sleep. We walk through the origin story of EnsoData, from the founders’ initial idea followed by 12 pivots to arrive at their flagship product, EnsoSleep. We talk about analyzing sleep data, solving hard data analysis problems with AI, and boostrapping enough data into the EnsoData ecosystem to build and validate the product.
We get into the long-term vision for the company, and even the potential for transferability of their algorithms and AI into other areas outside of sleep.
Comments or suggestions? Email us Feedback@TheMebFaberShow.com or call us to leave a voicemail at 323 834 9159
Interested in sponsoring an episode? Email Justin at firstname.lastname@example.org
Links from the Episode:
- 0:40 – Intro
- 1:32 – Welcome to our guest, Chris Fernandez
- 8:10 – When the focus shifted to sleep
- 13:31 – Collecting preliminary data
- 16:57 – After Y Combinator
- 20:30 – How the clinicians responded to their data
- 22:50 – Going to market
- 25:17 – An overview of the sleep clinic space and its evolution
- 27:41 – Treating apnea
- 29:02 – Wearable tech for sleep apnea
- 31:16 – Growth plans
- 34:13 – Raising money
- 37:16 – What’s on the horizon
- 38:41 – Translating into other areas
- 41:09 – A look at the company staffing
- 42:11 – Best lessons for sleep
- 44:36 – Other places where AI can help in healthcare
- 46:14 – Most memorable moment of the startup journey
- 47:12 – Why We Sleep: Unlocking the Power of Sleep and Dreams (Walker)
- 47:15 – The Sleep Revolution: Transforming Your Life, One Night at a Time (Huffington)
- 49:07 – Where they can learn more: https://www.ensodata.com/
Transcript of Episode 239:
Welcome Message: Welcome to the “Meb Faber Show,” where the focus is on helping you grow and preserve your wealth. Join us as we discuss the craft of investing and uncover new and profitable ideas, all to help you grow wealthier and wiser. Better investing starts here.
Disclaimer: Meb Faber is the co-founder and chief investment officer at Cambria Investment Management. Due to industry regulations, he will not discuss any of Cambria’s funds on this podcast. All opinions expressed by podcast participants are solely their own opinions and do not reflect the opinion of Cambria Investment Management or its affiliates. For more information, visit cambriainvestments.com.
Meb: Hey, podcast listeners. No sleeping on today’s show. Our guest is co-founder and CEO of EnsoData, a startup using AI and machine learning to analyze waveform data starting in the sleep space to save clinicians time on labour-intensive, complex, data interpretation and help them across the healthcare continuum. Today’s episode, we drink some Red Bulls and talk all things sleep. We walk through the origin story of EnsoData from the founder’s initial idea, followed by 12 pivots to arrive at their flagship product into sleep. We talk about analyzing sleep data, solving hard data analysis problems with AI, and bootstrapping enough data into data ecosystem to build and validate the product. We get into the long-term vision for the company, and even the potential for transferability of their algorithms in AI into other areas outside of sleep. Please enjoy this episode with EnsoData’s, Chris Fernandez. Chris, welcome to the show.
Chris: Thank you for having me. Excited to be here.
Meb: Where in the world are you right now?
Chris: I’m in Madison, Wisconsin right now, as is most of our team.
Meb: Supposed to be a world-class university town, I’ve never been. My background today, even though I’m in my guest bedroom is the Manhattan Beach Pier. They just announced in Los Angeles, they’re closing the beaches for the July 4th weekend, which I’m incredibly sad about. I’m actually getting ready to hit the road. We’re heading to see some family in Colorado, in Utah, and Wyoming. So listeners, if you wanna do, like, a 10-foot meetup, shoot me a message, and we can grab a beer somewhere. And it’s gonna be a lot of fun today. How’d you sleep last night?
Chris: I did get a good night of sleep last night, ironically. That’s not the most common thing for me or even sleep clinicians, actually.
Meb: Well, we’re gonna talk about all things sleep today. My sleep is largely determined by… I’m the world’s easiest sleeper. My wife is on the opposite end of the spectrum. We kind of clash and meet somewhere in the middle. But as I was thinking about this podcast, my brother… I mean, there’s at least three or four people in my family that probably have sleep apnea, whether they know it or not. But I wanna hear the origin story, founder of, let’s hear the beginnings,
Chris: So we wanna take it way back. I initially, you know, met my co-founders at the University of Wisconsin, Madison back in 2010. We were sort of the nerds that fell in love with the confluence of applied artificial intelligence, massive datasets, and working with really great clinicians. Our CTNI had initially started a custom apparel company that had one of the first online marketplaces that was integrated into Facebook. We grew that thing really fast, got it up to 60K in revenues in our first year, while we were sophomores in school, ended up selling that business. It wasn’t a super exciting exit, but it was a cool experience. Then we looked at each other. We were like, “Why are we studying biomedical engineering and selling t-shirts?” Probably try to do something that’s closer to that strike zone. And then we got very lucky. The genesis of EnsoData kind of came out of a course called biomedical engineering student design. And every year, people get assigned projects. And so there’s a huge list of projects. I made my top 10. And then there’s a raffle and kind of what order you get to choose it. I was last, of course, out of hundreds of people. And all my top 10 got taken. And there was, like, one project left. Like, I guess we’re gonna do wireless pulse oximetry. We picked it and then the professor said, “Ha, the hardest project, like, all right, here we go.”
But we got really lucky. We had an amazing mentor and advisor, Dr. Fred Robinson, who’d initially sponsored that project, read, kind of prior to being an anesthesiology professor at the University, was previously the CEO of Marquette Medical Systems, helped to sell that to GE, served as their chief clinical officer, and then went on to serve as CEO of TomoTherapy through their IPO and eventual exit to Accuray. So we had somebody that was phenomenally experienced in medicine, venture capital, startups, and kind of running a business. And I think that he saw that we had a level of determination that represented, “We’re gonna do this, whether or not he helped us and we’re gonna bang our heads a lot in the process.” And so I think that he convinced himself that he needed to save us some pain and became one of our earliest kind of mentors and advisors. That company or that project ultimately went through more than a half dozen pivots. And I could walk you through those if that would be of interest. But sort of along the way, it may be helpful to walk through the pivots to contextualize what we do. What we initially built was a device that would wirelessly measure pulse oximeter data, so blood oxygen content and heart rate, would measure it from your earlobe, and it used advanced multimodal wireless systems to push that data up to the cloud through a combination of cellular, Wi-Fi and Bluetooth networks, and utilize powerful machine learning in the cloud, try to help keep congestive heart failure patients and other complex cardiac patients out of the hospital.
In order to launch a hardware business, especially an FDA regulated hardware business, that requires upfront capital. And just to prove that our product worked, we needed several hundred thousand dollars. We were ready to drop out. We got real close to securing some venture capital, you know, back in 2012. But ultimately folks advised that if you wanna do a business like this in healthcare, it’s generally nice to have, to some kind of a degree, you should probably finish that, he said. So we stuck around, found the university to be an amazing sort of launchpad for startups, I would recommend that to anybody as possibly the best place to start a company. You have an excuse to not know anything. You’re there to learn, and it really created an environment where a lot of people kind of lent their expertise and helped us to figure out what we needed to do. So, we couldn’t raise money to do the hardware business. What do we do? We pivot software only. Let’s do a business that we can take all the way to market without any capital because that’s the situation that we’re in. And so we pivoted to trying to sell that technology to other medical device companies. And the initial response, they were really excited, the digital blood pressure companies, the companies that made home sleep apnea testing devices, they said, “Awesome. We wanna go wireless. We wanna…” IoT was a big thing back then. “We wanna get connected. We wanna use this powerful machine learning. What are we gonna do with it?” And we said, “I don’t know. You’ve been in this business for 20 to 30 years, right? Don’t you know better than we?” That was a lesson. The lesson was we need a killer application. We were a solution looking for a problem. We built this thing that we validated good and just the same amount of data as Netflix as cloud. We broke the internet, or a part of it in the process. That’s a separate story. That’s what we needed to do. So we set out looking for a killer application. We interviewed more than 100 different physicians from every specialty of medicine, haematology, hepatology, endocrinology, all the fancy and esoteric specialities, and found an amazing problem, and an opportunity in the world of sleep. It was very clear to us as soon as we found it, we set out.
Meb: It’s funny, I was listening to tell that story, I mean, there’s so much randomness, serendipity with just life, and entrepreneurship, and hearing all the twists and turns. I was a fellow biomedical engineer. You can see my cup which randomly I have today. I graduated a few years before you. All right. So you guys kind of are working your way through the problem, trying to find this product-market fit. You eventually settle on this area of sleep. What’s the kind of next progression? Were you still in school at this point? Had you graduated?
Chris: We were still in school at that point. You know, we transitioned from undergrad to grad. I’m glad that we took the advice to stay in school because when we graduated, I felt an inch deep and a mile wide. Biomedical engineering is incredibly broad. You take courses in chemistry, and physics, and biology, and statistics, and mathematics, and every other engineering discipline, from mechanical engineering, to electrical engineering, computer engineering. So you get to taste everything. And it’s an excellent kind of fun foundation of knowledge from which to build from. But we realized that if wanna have major strengths, we wanna develop real expertise, we needed to kind of go further. So our CTO and I decided to go into the Master’s degree programs, the graduate program for Biomedical Engineering. I got heavily schooled in at the end of it because my VME masters ended up being mine computer science classes, right? I don’t know if you can call it… We’re like, “Trust us.” There is a really important biomedical engineering application at the interface of computer science and engineering. And since then, they’ve actually created a Master’s degree specifically in machine learning for signal processing. Within all the different places we could focus, we had this inkling, this is 2014, that AI was going to absolutely transform the way the world worked.
You know, we had several technological revolutions leading up to it and enabling it. We saw the microprocessor try to come out around the ’50s, and we got personal computers. Those got connected to the internet. In the ’90s, during the 2000s, we saw the proliferation of smartphone technology and cloud computing that produced a massive deluge of data that had to be dealt with. And now we live in a world with, you know, more data than we conceived that we could even store a decade ago. And, you know, we look at AI as really bringing those technologies to life. It is a way to solve increasingly complex problems that get into the realm of tasks that require expert human level decision-making, problem-solving, reasoning and expertise to be able to do effectively. But to bring these systems to life, it has opened up opportunities to try to solve a whole new category of problems that was sort of off-limits due to the power of those prior generations of methods. And so, decided to focus on AI, took all the AI classes they had. Went up through the 800s and 900s in statistics courses. Got my doctorate plenty in stats courses but learned a lot. And within AI, there were incredible breakthroughs over the last several years in image recognition and computer vision, and natural language processing, and speech translation, and in a variety of other applications of AI.
But there was one particular area that felt like it was at least five years behind the progress that those other kind of subdomains were making. And that was the application of AI to waveform. And so, you know, even to something like an image, waveform data is heavily unstructured. It’s arbitrary in length. There’s different sampling rates and resolutions. There’s tons of different kinds of sensors. But what we end up with is this really is the most widely available data for diagnosing and for treating patients. We have a proliferation. Nearly every adult in the world are connected to the internet. You know, an asymptote to zero cost of the sensors that are used. And more than a billion waveform studies run per year to diagnose, monitor and treat patients in nearly every speciality of medicine. And we’re like, “Why is nobody working on this?” And so that was sort of the toolset, the framework that we used when we went looking for that killer application. And we saw an opportunity by applying waveforms to dramatically improve patient access, patient outcomes, the in-patient provider experience and the affordability of healthcare. So that was our mission from the beginning. And when we looked at all these different conditions, we saw a really unique opportunity. And what’s so unique about sleep is that well, one, we all sleep. Researchers estimate there’s nearly 1 billion patients worldwide that suffer from sleep apnea. Today, they’re 80% to 90% undiagnosed. Researchers estimate there’s more than 50 million patients in the U.S. also over 80% undiagnosed, and particularly in critical access, rural and underserved communities, you know, under-diagnosis can approach 95% or more. You know, we looked at things like epilepsy and looked at things like, you know, AFib and arrhythmia. And there’s opportunity to make a huge impact in those medical specialities. When you find something, when there’s a billion people that have it, and they don’t know yet, that felt like a really unique opportunity to try to massively scale up patient access to information about how they sleep, how to do it better, and, you know, how it affects them.
Meb: So you guys have this thesis, and where did you kind of start next with acquiring the data or starting to build, you know, going through the scientific method, say, “Can we apply these tools, specific domain?” Would this have been about what year? 2014, 2015, 2016? Where are we now on the timeline?
Chris: So, now, we’re in 2014. There’s two parts to that answer. The first part is, you know when you know. So relative to the other interviews, the other interviews are interesting, but the sleep interviews we had literally what the medical director said to us, the first institution we partnered with is, “Oh my God. Come with me. Here’s the kind of the bunk where we do all this work. If you could solve this problem, be broadly applicable, be applicable immediately, be applicable to every sleep clinic in the U.S. and globally, and I want it so bad, I’ll give you 3,000 patients worth of data to see if it works.” And so people are literally trying to give us their data. And, you know, in healthcare, there are a lot of complexities around utilizing data for research and other things. And so that was an early signal that this was a good idea. I think the harder question though is, every AI company, typically in healthcare, has to solve a data chicken or egg problem, which is you need data to build a product that creates value. And in the typical voluntary exchange of value that people like, you don’t have anything you can give people. Really at that first stage was validating the market and validating the demand through selling the vision. We called every single sleep clinic in the Midwest to do that, talked to them on the phone. They all said that this was a thing. We said good.
We went through… So I guess after that, we were sort of at a crossroads between staying in graduate school or do we cut and run. And my academic advisor told me at that time, “You’re doing too many things, you need to pick.” And a couple of days later, I had gotten to Y Combinators, first fellowship program, where they picked 35 companies out of more than 7,000 applicants to move out to the Bay Area and do that whole thing. We said, “All right, we got to YC,” set sail from there. At YC, we drove to every single sleep clinic within 50 miles of San Francisco. And it was simple. It was, “This is our vision. This is our mission. We don’t have anything we can give you. We need data. If you want to support this and see if we can even have a shot at making it happen, kind of that’s what we need to do. And if it does work, we’re still gonna need to go get an FDA clearance afterwards. So it’s not gonna be ready soon.” We were able to go out and sell the vision. We were able to bootstrap enough data into our ecosystem to be able to build the product, to be able to validate it clinically, and ultimately, kind of take that to market.
Meb: What’s the sort of minimum number of data points with the patients you would need, and then at what point do you start to do FDA submissions or get it approved? How does that all work?
Chris: You know, I think from an FDA standpoint, it’s different for each different kind of product. Each product has its own unique sort of sample size and fiscal power that go into that design. But to do really fancy machine learning and deep learning, you’re really talking about terabytes, and petabytes, exabytes scale data. So you wanna be working in that domain. You know, sleep studies can range from anywhere between a half gigabyte to five gigabytes each. Even at the starting point, we were in the 100 gig range, and then quickly, we got into the hundred terabyte range for that.
Meb: You come out of Y Combinator, you talk a bunch of sleep centres into giving you some data. What’s next?
Chris: What’s next? We come back from Y Combinator, we’re pre-funding. At that point, my co-founder and I are running the company on our credit cards. The goal at that point was to get funding. We had validated the market demand. We had enough data to build the product. We were ready to go, kind of do the FDA thing and that required capital. So that was really a crossroads. We said, “We’re living…” Our co-founder and I were living in a dingy apartment that we had negotiated a month to month rent on, and I was getting close to maxing out the credit card. So he said, “When are we gonna pull the plug?” “We’re gonna pull the plug when I actually can’t even buy a bus ticket home,” call my parents and say, “All right, it’s over. We’re gonna go get real jobs.” We’re just playing chicken against the end of our own personal financial runway. We were really lucky. The day that we got back, it felt like kind of putting out feelers in the fall of 2015. And then the day after New Year’s, we met with an amazing digital health investment firm, HealthX Ventures, Mark Bakken there. And he immediately appreciated our vision for data science, how valuable the application for machine learning and AI was in healthcare and with waveforms, and really how rare and unique our company was with our level of scientific sophistication data. And I don’t think we’re bad at telling you, they put together our first round of seed financing. We did a half-million dollars on seriously financing in early 2016.
The first thing that we did was hired our CTO back who couldn’t take the initial leg of the journey with us, but we were jonesing to bring him back the whole time. And then the three of us took on the FDA submission. You know, it was something that we didn’t have, you know, a ton of direct experience with. We put together a team of very experienced advisors and consultants to help kind of guide us and point us in the right direction. But we recognised that if we really wanted to build a company applying AI to the heart of diagnosing and treating patients, that having, you know, regulatory competencies, both in terms of FDA and in terms of data privacy was really something that needed to be a core competency. And so, we could either look at it like, you know, that’s a lot of work… Like, it seems a lot of people do. Or this is actually something that can create some competitive distancing and some space forward, behind and side to side. Why don’t we embrace this opportunity, try to do it better than everybody else, try to innovate inside of it, and try to use this to kind of push it forward faster, and to create differentiation? So that was the approach that we took. We wrote the documents internally. I wrote 1,000 pages of software documentation that summer. We went on one vacation. I remember we did a road trip to Denver to go see a concert, and to hit the FDA deadlines, I was in the backseat the whole time typing. Coding and writing in the car was a very common occurrence for us at the early stages as we parallel across those words.
Meb: So, do you guys go into Red Rocks or where?
Chris: That was the Dick’s Sporting Goods arena that time. Yeah.
Meb: My family lives sort of near Morrison. So, I’ve spent many an evening at Red Rocks. All right. So you guys eventually finished the FDA approval process. Where would we be now?
Chris: We ultimately got our FDA 510K clearance to aid clinicians in diagnosing sleep apnea and other sleep disorders by automating the analysis of the physiologic waveform data that came from sleep studies.
Meb: And walk us through kind of how the clinicians approached it, either before or without your assistance.
Chris: Pre-answer data, you know, more than 80%, 85% of the sleep clinics out there today in 2020, manually by hand score all of their sleep studies. So what does that mean? What does that look like? Some clinicians will pull up the sleep study. And it can contain 20 sensors or more. It includes brainwaves to measure sleep quality and how disruptive fragmented sleep can be. It includes two kinds of airflow to be able to measure breathing and breathing frequency. That includes respiratory effort on the chest. It includes an electrocardiogram for heart rhythms. And it includes pulse oximeter for blood oxygen saturation. So you’re talking about dozens of different kinds of sensors. On those sensors, they are reading these this data in 32nd pages. And on each page, they’re using visual pattern recognition and perceptual reasoning to identify in the circle and to annotate, and to mark, and to tag different kinds of physiologic patterns. And that includes things like five different kinds of sleep stages, apneas, hypopneas, blood oxygen, saturations, neurological arousals, bradycardia, tachycardia. I could keep going. So, it’s dozens of sensors, dozens of different types of complex events. And, you know, when you’re talking about sleep apnea, ultimately, you’re talking about hundreds of thousands of events per patient that need to be manually done. So on average, that takes about an hour or an hour for an in-lab sleep study. And we’ve taken an approach to try to, not to try to but to leverage artificial intelligence to do that analysis automatically, and to achieve such a high level of quality that clinicians can really establish a level of comfort and confidence in that analysis to be able to go through it much more quickly than they’d otherwise be able to. And in a speciality, there are so many undiagnosed patients out there that need help and can really improve their quality of life and their health by getting access to these services. The real opportunity is trying to save these clinicians as much time as possible so that they can actually increase that access.
Meb: It’s such an obvious why in the world any clinician would be spending all this time reviewing these mounds of data. It seems so antiquated. You’ve identified it works, you get the FDA approval, and then kind of what’s the next sort of step in y’all’s evolution? Is it just to start hitting the road, chatting up a bunch of sleep clinics? How’s it work?
Chris: Go to market, sales. That point we still had three people. My other two co-founders are technical as well in background. And so once we got the FDA approval, got the FDA permits, we had some pineapple orange juice mimosas. That lasted for like 30 minutes. And then I think we’re workaholics, we’re like, “What do we do?” We just started working again. And, you know, we got to it. But initially, the go-to-market, the three of us sat down and were like, “All right, now we can sell it. Who’s gonna do sales?” And they both just point at me like, “You.” So I was unanimously elected to do sales. They held on the technical for it. And we then hired our COO, Brock from Epic systems. He had, you know, five years experience working with the largest enterprise health systems and academic medical centres in the world, implementing software and other IoT solutions at a very large scale. So that was just the kind of skill set that we did to be able to grease the skids and to really streamline our launch in the market. And we went to it. We utilized… I think one thing that we do is unique is having a very sort of scientific, rigorous medical approach, but also trying to take best practices from tech and from, like, the YC type world, and apply them in a really interesting way at that interface. So, we decided to leverage a free trial model, like many or most SaaS solutions do to drive customer acquisition. Let’s just try to get this thing out there for free because seeing is believing. People have been trying to automate this for the last 20 or 30 years. Nobody has cracked the nut. Everybody who’s tried it has had a horrible experience, flatly they don’t believe it back in 2015. And I think there’s been a lot of evolution in the public, you know, just with all the dialogue there and all the news on artificial intelligence, that’s been very influential in sort of the evolution of adoption in a clinician’s mindsets. So use the free trial model. We started going. We got to around three live customers with around six sleep clinics by the end of 2017. And then it’s been scaling since then. Off to the races, now we’re working with more than 300 sleep clinics around the U.S., some of the leading academic medical centres enterprise health systems, sleep clinics, integrated diagnostic, testing facilities, and increasingly portfolio partnerships with medical device informant.
Meb: As most of the sleep clinics you go to, right, it’s in-person where you go and get tested. Is there also any sort of at-home sort of diagnostic or is there any sort of kits on the horizon where people could self-test at all or what’s the lay of the land for sleep clinic base, in general?
Chris: Really interesting. The, you know, home sleep testing technology got validated several decades ago. And then in the 2000s, we started seeing adoption. In the last 10 years, U.S. homes and testing volumes have increased by more than Pontiacs. In the last two years, they’ve doubled. And main drivers in that adoption of home testing are, you know, a lot of patients feels more comfortable. They don’t wanna sleep, you know, in a hospital or in a clinic and have people kind of supervising them. They like the comfort of their own home. Medicare and Medicaid reimbursements, it’s only one third the cost to do it at home, as it would be in clinic setting. So it’s more affordable. It’s more comfortable. That has created patient and payer preferences around that. And we’re seeing a lot of the growth in the sleep medicine industry, occurring in that push of patient care into the home setting in the home environment, both through home testing devices, through cloud-connected see pap devices, and through telehealth. I think that we’ve seen what, you know, somewhere between years, probably years of the trend in the shift towards Thompson testing, which we felt was inevitable, get compressed into a span of days and weeks during COVID. And during COVID, many sleep labs that had a discontinuity in service. Their in-clinic environment was disrupted by social distancing and by CDC guidelines, in some cases. And so, the clinics that were already well-equipped and at scale in-home testing, many of them [inaudible 00:27:08]. And we’re able to really maintain continuity and patient access to these important, you know, sleep services during these hard times. And a lot of people had tried to adapt quickly. But today we do live in a world where effectively all sleep medicine care can be done virtually. And from our perspective, we as a field should put the right practices in place so that this pandemic or the inevitable next pandemic, whenever that does happen, is not something that causes patients to lose access to healthcare because it simply doesn’t need to .
Meb: What’s the current gold standard for apnea? Is it just the oxygen mask at night or is there anything kind of being developed? What is the actual treatment?
Chris: So typically, first line of defense for a patient that has obstructive sleep apnea would be CPAP therapy, which is continuous positive airway pressure therapy. You know, there are more advanced versions of PAP therapy, as those conditions get more severe and complex. There are alternatives as well. You know, it’s a little hard, but one of the most effective ways to lose 10% of your body weight, in many cases, that has a meaningful impact on the severity of the sleep apnea but we lost as hard. Other, you know, options and alternatives include oral appliances to sort of create the right spacing in the mouth for that breathing to happen. There are a new category of neuromodulation and other implantable devices that can provide an internal implantable versus an excellent device sort of solution. There are surgical procedures as well. But overall, you know, CPAP compliance and CPAP adherence in the United States and globally is very low. And I think that that’s a signal that there’s a real desire among patients for an interest in, you know, new possibilities and new categories of therapy that may be more comfortable, more tolerable, more enjoyable to do over the long-term.
Meb: I imagine people are listening to this, they’re talking about a lot of the wearables like aura rings with the NBA, going potentially back to playing games this summer, about being able to try to keep an eye on COVID. How far are we away from…? I mean, this seems like a pretty involved test. You mentioned with the EKG, and the pulse oximeter, and everything. How far are we away from a much more sort of simple test or even a wearable in able to detect and diagnose something like sleep apnea? Is that on the horizon or is it like 20 years?
Chris: I think it’s a lot closer than most people probably think it is. We do envision wearables as being an important mechanism for patients and consumers to be able to access information about how they sleep and how they can improve it. I think the great thing about wearables is, you know, sleep is really not just a thing for patients. It’s really a thing for all of us. And if we have opportunities to do it better, it can create a really perceivable improvement in your quality of life. It can be multi-generational in the sense. If you imagine really tired parents and their ability to sort of, you know, help all their kids in various ways, you know, the impact that you can have really isn’t just even limited to one person, but it can impact positively a lot of people in their environment, in their community as well. So we think wearables are a really important and awesome kind of trend that are getting information about seeping into a much greater number of people’s hands. And we’ve been partnering with leading wearable companies like Veteran, Bio, IntelliSense, and bring clear, and others to be able to help them get their solutions to market faster at a lower cost, and really to allow them to focus on their own competitive differentiation. There’s an entire wave of interesting novel kind of wearable technologies coming in. We envision ourselves as sort of the digital health infrastructure that can analyze their data so well, that it turns out something like MongoDB, where, yeah, you could code your own database. But why would you do that? Those companies need to win against each other. And I think that comes in how comfortable is the device? How high quality is the data? What novel sensors can we add to it? How do we wrap that in a nice kind of patient experience?
Meb: This is how this product that seems obvious in a lot of what you’re doing with the sort of just blocking and tackling of scaling, and partnering, and growing revenue, that’s all going on. But you guys have your hands involved in a lot of other different ideas and stuff. Can you walk through kind of where we are here in 2020, what other fun expansions either this analysis or other things you guys are looking at to the extent you can talk about it?
Chris: So I think that we have some very exciting new products coming on the pike. So look out for those. But in terms of where things go from here, you know, outside of our core products, our peer-reviewed research… Peer-reviewed research is something that we feel is very important. We have continued to do that. We’ve published more than 12 peer-reviewed papers on the interface of AI and healthcare, 2015. We’ve done that to try to advance the state of knowledge and just contribute to pushing the field forward to better understand how sleep influences human health and disease. And what we’ve shown through that research is that it’s not just analyzing these diagnostic studies. But AI is really going to impact every step of the patient care journey, starting from being able to better identify and screen for undiagnosed patients in much easier and more convenient ways using things like wearables and electronic medical records, being able to prioritize patients based on who’s really the sickest here and who’s gonna have the most negative impacts, you know, if the diagnosis and treatment isn’t prioritized at that sort of level. You know, we’re already optimizing the time and the cost that’s involved with diagnosing patients, and then on the back end, really being able to try to bring personalized and precision medicine into a reality.
I think there’s an increasing awareness, you know, in the field that sleep apnea is not just sleep apnea. There are actually a lot of different factors or reasons why a patient could have sleep apnea. It could be anatomical. You know, it could be related to their BMI and sort of the potency, they call it up their airway. And you can imagine that based on why they have sleep apnea, some treatments may be especially effective. And some treatments may just not kind of totally circumvent the underlying reason, the underlying causes. So there’s a big push in what they call phenotyping or endotyping diseases, to take them to the next level of resolution, to be able to understand what individual or combinations of factors are most likely causing those diseases and being able to take a much more targeted and personalized approach to, you know, really trying to address the root cause. That’s an area that we’re really excited about. And, you know, of course, as we said with the current kind of compliance levels on CPAP, we’re really trying to support pharma companies and medical device companies in bringing forward that next generation of more patient-friendly, more comfortable treatment options that folks have more to choose from.
Meb: Started the company with racking up a lot of frequent flyer miles on credit cards. You guys did an early raise and you guys just raised another big slug. Do you wanna tell us about that process was a little easier, this go around?
Chris: It was interesting to go around. I would say I learned a lot of lessons. But I think we were really happy with how it ultimately came together, sort of kicked off our fundraise at the JP Morgan healthcare conference in San Francisco. That’s at the beginning of the year in January. My recommendation to any and every healthcare founder is raise there because it’s the beginning of the year, everybody’s sort of like, they got the fresh books. They’re ready to do the first deal of the year. Everybody’s looking for partnerships, and it’s kind of a circus. It’s a circus, absolute circus, but it’s a circus where everybody’s all at one time. You know, leaders IN the medical device world, the pharma world, VC, you know, plays big and healthcare, hedge funds and other institutional investors. And so it’s excellent. We ended up having, I wanna say, audio meetings in three days or four days. it was the busiest couple days ever. And that really kind of helped us to accelerate that fundraise process. You know, we were initially arguing to raise about 5 million. we ended up getting quite a multiple of that in commitments in terms of our subscription. So at the end of it, we really had the good fortune of being able to pick out a syndicate of Venture Capital Partners that had real domain expertise and unique value to bring to the table. We were able to raise so we close twice as much money as we set out to raise we closed $9 million, and then included Zetta Venture Partners, one of the best AI-focused kind of seed-stage funds that’s out there with great expertise in healthcare. That was co-led by venture investors as well. They have broad expertise in the healthcare industry in these different segments. SleepScore Ventures was also in the round, one of the first sleep-focused venture capital funds that has a lot of amazing strategic relationships, and others, you know, including Dreamit Ventures, a VC fund associated with one of the accelerators that we did, which is awesome. I highly recommend Dreamit as well, and Necessary Ventures, HealthX Ventures, Fully Capitalist. So it was a good round.
Meb: So what do you guys gonna spend all that money on?
Chris: It’s really simple. Ramp revenue, launch new products, and continue building out our amazing team. So we’re doing a little bolus of hiring right now just to increase our firepower and our capacity. And using that capacity to focus on wanting to continuing to expand our partnerships with our core customers and others in the industry, as well as bringing new products to market that are really going to change things.
Meb: I used to go to the JP Morgan conference every year so I actually started out as a biotech equity analyst in the depths of the last big internet biotech busts. So I was going to that conference over the years have long moved on to being a quant. I’m a tiny investor in this last round of years, last year, oddly enough to buy about a round of what you’d say? Pineapple mimosas. And that’s about it. But I’m cheering for you guys. Are there any sort of expansion ideas that you can talk about, that you guys are thinking about or things that you’re brainstorming, you got your eyes on the horizon about or are these all under lock and key? Anything you can talk about?
Chris: There’s a few different directions. Like, I shared a little bit on, you know, within sleep specifically, where else outside of the diagnosis? You know, AI is contributing to helping clinicians to increase access, improve outcomes, and address affordability of care. But I think, you know, one of the big expansion opportunities is going to be on sleep. These waveform studies are used in nearly every specialty of medicine. It includes things like eg studies, or that are used to diagnose epilepsy and other neurodevelopmental and their general conditions that includes, you know, monitoring devices that are used in surgical procedures, and anesthesiology. And in critical care unit, it includes electrocardiograms that are used to diagnose cardiac conditions. And so, I played this game, if we wanna try to find the specialty of medicine that doesn’t use waveforms for something important, I’m able to find one yet. And so really, I think that we’ve done the initial work to validate that our AI is ready to address some very important problems outside of sleep. And, you know, we feel like it’d be a disservice to the world to kind of not try to share and to contribute, you know, all those power capabilities, you know, in other areas as well.
Meb: How portable is that? So say you have your engine that you guys have built, how simple…? And I’m sure that’s the wrong word. But how difficult is it to say, “Okay, we’re gonna look at this totally different data set. Is it something that build-out would take a month or would it take like two years?” Like, how complicated is the intransferrable as the work you’ve done into a totally different area?
Chris: So, transferable is the magic word. I think that’s the art and the science. One of my favorite research papers on artificial intelligence and healthcare was done by researchers at the University of Chicago. And in that paper, they wanted to develop an AI system that could, you know, detect cancer, and tumors, and breast mammography images. So the standard imaging that’s done for breast cancer. Well, fortunately, they wanted to use deep learning, which is really powerful. But unfortunately, you can’t get very far if you have on the order of a few hundred data examples to be able to train that system. And so they are really limited by just the amount of that mammography data that they had. What they ended up doing was, they took the image net dataset, which is, you know, more than 14 million images, that include more than 20,000 different human-annotated objects, like cats, and trees, and airplanes, and dogs, and stuff like that. Then they trained the algorithm on the image mapping data set. And then they did this procedure that we called transfer learning to try to use what the AI had learned about pulling out useful information, structures, regularities, patterns out of that format of imaging data, and to just try to generalize and borrow some of those learnings to apply it so that they could juice up performance on this much smaller data set. They were able to achieve breakthroughs state of the art performance and ultimately, you know, detecting and categorising cancer in breast images. And so I think that’s just an example of some of the categories and tools that are available in machine learning and AI that aim to address that problem directly. And so, part of the transferability as a resource allocation, how many people can we fork off our core businesses are working on this other stuff? But there are machine learning and AI methods that we are and have been developing that directly solve that problem of transferability. We’ve done some initial validation of our software in several areas outside. I can’t go into a ton of detail in just …
Meb: It’s all right we’ll have you back next year or whenever and you can chat about all these things. What’s the headcount out now? How many folks y’all got?
Chris: We got 15 now. We made two offers last week. We made one offer this week. So, you know, I think that it is a good time to be hiring. I think there’s a lot of really talented people out there that just purely do the external environment of COVID, have been put into place where they’re either out looking for work or the place that they’re working, it’s become, you know, a little bit less enjoyable. So I think that we feel really, really lucky to have been able to, you know, secure this financing during the pandemic, during tremendous economic uncertainty. I think that speaks to the commitment and conviction of our investors, how important our mission and our vision is. And, you know, whether or not there’s economic downturns, whether or not there’s pandemics, this needs to happen. But that’s allowed us to be able to look at increasing our resources. And we’ve had, I wanna say, maybe 100X or more increase in inbound applicants on a monthly basis.
Meb: Okay. Listeners, you heard him, shoot over your resume as you hear this show. You know, it’s fascinating to me that you mentioned that there’s a straight-up sleep focused fund, which is pretty incredible. What have you learned about sleep in your own life for over the past 10 years? You got any…? What’s your hacks? I’m a cold guy and held out, I haven’t bought one yet. So maybe you can push me in that direction. I’ve always wanted to do one of the chili pads or now they have, I think it’s like $2,000 eat sleep sort of setups because I like it super cold. My wife doesn’t. What else you learned about sleep that might be helpful for our listeners?
Chris: I can share the single biggest and most important lesson that I’ve learned about sleep personally in my life, and I learned that lesson during college before we were playing work on sleep, back, if you pulled me out of the time machine and you asked me if I thought that sleep is gonna be fine. It’s not. You know, in the first half of college, let’s call it, I was frequently staying up until 2:00 am to try to study for tests, stay on top of deadlines and projects. Often that would turn into 4:00 am, sometimes it would turn into an all-nighter. And I found myself really tired a lot, and all that, and tried to do an experiment which was feels like I’m really not being as productive as I could be. I wonder what it would feel like if I just, like, slept an adequate amount, like seven or eight hours. Like, what would I be able to do? So I decided to force myself to do that, and try to go to sleep by midnight, and try to wake up 7:00, 8:00, 9:00 am every day, establish a nice routine. That’s basically I worked less. I studied less, I slept more. I got much better grades, I was much more productive and I was much happier. And that was like the needle that seemed impossible. It’s like, how are you gonna spend more time sleeping and then how are you gonna account for that on the other side? It was really amazing to see that. And that was a much better, more sustainable, and more productive way to approach things. So I’d recommend everybody try to just sleep adequately and playing that other game. There are real limitation…
Meb: One of my buddies jokes, he says, billion-dollar idea, so listeners, feel free to steal it is he wants to build a mechanical bed so you can start the night with your partner cuddling and then it separates as you both fall asleep so you get to sleep alone. So they’re not harassing you, elbowing you in the head, snoring and then by morning time, it comes back together. So there you go, listeners, feel free to take that one. Well, as we start to wind down, I got a couple of fun questions, but anything in particular you think is on your brain, you’re excited about. You obviously have a curious mind as we begin this decade in 2020.
Chris: The thing I would say is I think it’s so important for startup companies to be driven by real purpose and mission. I think we’ve really been able to leverage that in a variety of different ways from solving that data, chicken or egg problem to recruiting, you know, some of the best people in the world. But, you know, I say that it’s 2020, and healthcare has really not benefited, you know, significantly yet from the AI revolution. Eighty-five percent of sleep clinics still manually go through all that data today. State of the art senior software tends to correctly identify zero seizures and about half the patients it’s run on according to an independent study in the American Academy of Neurology journal. UCSF false alarm study showed that 90% of the alarms that went off in cardiac ICU were false alarms. And nearly 100% of the data that’s generated in healthcare, siloed both inside of different vendors, proprietary systems, and inside of the medical specialties that don’t communicate a whole lot. So, you know, we’re really trying to break down those barriers with respect away from AI, specifically… But I think there is a time, there are so many other applications for AI in healthcare. And for all the folks out there that are AI people, you know, it’s a lot more fun to apply AI to a mission like this, then do we put up the Bears jersey or the Packers jersey? Which one are we gonna sell on? You know, sell more toasters and umbrellas. It’s a totally different game. To do this in healthcare, I think it’s really rewarding. And I want more people to have that up too.
Meb: What’s been the most memorable moment of this entrepreneur journey of the past? Is it five, six, seven years now? Anything come to mind, good, bad in between?
Chris: The entire thing, really. Like, I think the biggest lesson I learned is kind of go for advice towards action. The things that you can do, like, I think pretty much everybody is capable of doing things that are probably way beyond what they think they’re capable of, things that seem impossible. things people say are impossible, but sort of just going for it is I think is a really big part of the battle. It is hard sleeping on couches. There’s a lot of, I’m gonna run out of money, all that stuff. You can weather that storm. As Paul Graham says, “If you don’t die, you’ll eventually win.” Stay in the game.
Meb: What was a really super popular, just trying to remember it, sleep book that came out in the last year or two? Do you know what I’m talking about?
Chris: There’s been a couple. You know, Matthew Walker at Berkeley wrote a book “Why We Sleep,” that was really popular. Arianna Huffington ended up writing a sleep book. And then it’s also got major coverage in “The New York Times,” “The Economist,” and “New Yorker,” and other outlets. So I think that that’s contributed significantly to public awareness about sleep, and the importance of sleep, and some of the things that happened to our bodies if we don’t sleep.
Meb: I’m not gonna hold you to this. But this is just the best guess. When we are… This is July 1st, we’re recording this, it seems like you’re a music fan. When is the next time you see a live show in-person? You gotta give me over under date. When do you think that’ll happen?
Chris: Well, you correctly categorised me. I’m a huge music fan, both listening to it, live, and recorded, and making it. We actually have a jam studio in our office. We have some very talented guitarist, pianist, drummers. We got a whole PA system set up in there. So for folks out there that like to jam, stop by any time, over under, not this year, unfortunately, really hopeful that by Q1 of next year, hopefully, early next year, we’re starting to have those opportunities. But I think that there’s a lot of interesting stuff happening in the music space right now. I know one group of artists I was following, put on a seven-week virtual Music Festival, and they had shows, like, all day, every day for seven weeks. I could barely watch any of it. But let’s create some really interesting opportunities for artists to try to, you know, promote their work and to share their work in new ways. And I think it’s been really exciting to kind of follow, you know, some of those developments, as well.
Meb: You got any favorites I can download Spotify later today? You got anything you’re listening to? Who are you listening to in 2020?
Chris: Recently, just because it dropped on Friday, I’ve been listening to the new bass nectar album, All Colours, nonstop since last Friday. I don’t know how many laps I’ve made through it yet, but more than 10. So I highly recommend it.
Meb: Awesome, Chris, people wanna find out more about what you guys are up to. They wanna follow along with the story. Where do they go?
Chris: Howdounsedated.com, we publish blogs. We have some really cool newsletters. And we’ve been ramping up our webinars as well. We just did our first one. And to my surprise, 750 people showed up. So we’re like, “I guess we want this. We can do all this.” And so there’s a lot of cool content we’re putting out through our website. I highly encourage folks to check us out there and to dig into our research more funds.
Meb: So a lot of fun hearing about this. Looking forward, follow along with your jersey. Thanks so much for joining us today.
Chris: You too, Meb. I really appreciate the opportunity, huge fan of the show and excited to continue listening and following along.
Meb: Podcast listeners, we’ll post show notes to today’s conversation at mebfaber.com/podcast. If you love the show, if you hate it, shoot us email@example.com. We love to read the reviews. Please review us on iTunes and subscribe to the show. Anywhere good podcasts are found. My current favorite is Breaker. Thanks for listening, friends, and good investing.