Reinventing Healthcare with IoT, AI and ML

What if there was a way to change the trajectory of human health? Our era is marked by the proliferation of inexpensive, but high quality IoT sensors that enable advanced early disease detection. A way to measure human physiology in a continuous manner.

That’s exactly what data sciences company Biofourmis is doing. CTO, Milan Shah explains how the combination of IoT, artificial intelligence, and machine learning is enabling clinicians to interpret patient’s biometric data and identify earlier warnings than ever before. This allows clinicians to intervene earlier, improve patients’ quality of life, and provide them with better outcomes.

Join us as we discuss:

• How Biofourmis is identifying physiological deterioration using ML and AI algorithms
• The types of medical-grade IoT sensors in the market today
• The power and potential of biomarkers
• How the COVID-19 outbreak was managed in Singapore with Biofourmis’ solution
• Why personalised predictive healthcare is the future

Subscribe to IoT Leaders

Subscribe now

Subscribe now

Subscribe now

Transcript

Nick Earle:
Hello and welcome to the IoT Leaders podcast with me, your host, Nick Earle, CEO of Eseye. This one I actually think is my favorite one. Of all the ones that we’ve done, I think this is my favorite. And the reason is, it is such … Well, it’s a great story, but this is a company Biofourmis, as you’ll hear, that is tackling an enormous problem, which essentially can be summarized as reinventing healthcare. And although that sounds hugely ambitious, and you think, “Oh, surely no one can take that on,” are they being funded? Almost half a billion dollars by some of the top VCs in the world, who believe in what they’re doing. They’re already active in the market, and they are essentially identifying disease using machine learning and AI algorithms, to identify disease before the patient sees it and therefore get early intervention. And you’ll hear a great story about their journey from Milan Shah, the CTO, and where they’re up to, and what else it could be used for going forward.

So this is a real Silicon Valley type play. They’re based in Boston, Massachusetts, although they started in Singapore. And so it’s all here in the podcast. I think you’re really going to enjoy it. And so with that, I will then hand over to my discussion with Milan Shah, the CTO of Biofourmis. Here we go.
So Milan, welcome to the IoT Leaders podcast. Great to have you.

Milan Shah:
Thank you for having me, Nick.

Nick Earle:
Well, really looking forward to this one, and we’ve been working together for a few years, and I like to say that Biofourmis is not a company that’s trying to solve a small problem. I mean, you are massively ambitious as a company, and quite a bit down the line. You’re the CTO of Biofourmis. Perhaps we can start by, maybe just a little bit about yourself, Milan. I always like to start off the podcast with just that people understand who my guest is. So maybe just a brief potted history of yourself and how you became CTO of Biofourmis.

Milan Shah:
Sure. Thank you again for giving me the opportunity to be here. It’s been a pretty amazing very, very … I consider myself very blessed to be able to do what I’m doing. So I’ll start with my graduate work, which I did at MIT. This was about 27 years ago. And at the time, I worked in the CSAIL, the Computer Science and AI Laboratory, and AI and machine learning were in a somewhat rudimentary stage. There were Cray computers, massively parallel computers being built to do AI/ML, and really the big thing that everybody was trying to figure out is, what would we use AI/ML for? What could be the applications of this kind of technology?

So that was 27 years ago. And since that time, my entire career has been really focused on both those elements, large scale computing and AI/ML. Now, my earlier part was more about large scale computing. Distributed computing was the implementation, not necessarily the supercomputers part, but the idea of distributed computing really was starting to take off and then the internet happened and now we have cloud computing. So distributed massive-scale computing is something that’s here and now, very easily available. My career journey started at Microsoft where I worked on the initial versions of Microsoft Windows NT, again on the distributed file system. I ended up building Exchange, which is their email product, still very much in use today. And then I moved on to a lot of cyber security type of applications, so a single sign-on and so on.

The recurring theme over there was, an underlying theme in all of these cases was, how can we also use some amount of machine learning to manage the scale, if you will? Certainly, the last thing I was doing in cyber security was essentially monitoring, collecting metrics from computing. So you are, let’s say, a large infrastructure company like a PayPal or an eBay and people like that. You’re running hundreds of thousands, or Facebook, you’re running hundreds of thousands of computers, and you’re trying to figure out which of them might be under attack, which ones of them might already be hacked and so on. So the general idea was, if we collect enough metrics from all of them and we understand what their baseline behavior is, maybe we can use AI and ML to identify which ones of them are skewing off the baseline and potentially have been hacked and so on.

It’s a tough business, tough technical problem, mostly because computers don’t really have a consistent baseline. And while I was working on this kind of a field, I managed to run into Kuldeep Rajput. He’s the CEO and founder of Biofourmis. And he had recently decided to move his headquarters to Biofourmis. This was on the heels of the series A, or series B I should say, from Sequoia, the 45 million. So he was ready to really expand the technical team and then take the company onto its full potential. He was moving his headquarters to Boston and that’s how I got introduced to him. And when I learned about what he’s trying to do, essentially collecting metrics from human beings and trying to figure out when somebody deviates off of a baseline, and therefore is likely deteriorating from a physiological perspective, if we could detect that and give clinicians an earlier warning than has ever been possible, then they could intervene earlier, produce better outcomes, the patient benefits from a significantly better quality of life, overall costs to the systems go down.

And we fundamentally can change the trajectory of human health. You can actually anticipate a day when, given systems like ours and automatic infusion systems, you could actually have self-medicating human beings, where for many, many cases, you could detect the onset of some illness and automatically administer medicine, keeping the human being essentially disease free to a lot of what is common diseases today.

So once I understood what he’s trying to do, the fact that humans’ physiology does actually, as opposed to computers, do actually have a very consistent baseline in many levels, I understood the power of being able to apply a AI/ML techniques to the problem of early detection of disease. And one thing led to another, and I decided to join Biofourmis, and here we are.

Nick Earle:
Here we are. Well, there’s a lot there, and it gives the listeners a good idea of why so many people think that this has a such huge potential. And to unpack that, I guess when I first heard of Biofourmis or I talked to other people about Biofourmis, they say, “Oh yeah, medical company,” but you don’t describe yourself as a medical company. In fact, you’ve talked about AI and ML, and when we spoke previously, you said, “No, no, no, we think of ourselves as data sciences,” and you talk about biomarkers. Maybe you could just go a little bit more into that, because it helps people differentiate between, is this just a big version … Are you competing with the Apple Watch or is this just a big version of remote patient monitoring? And it’s not, but to understand why it’s not, you have to go down the AI/ML, biomarker route, don’t you?

Milan Shah:
Absolutely, absolutely. So we certainly consider ourselves a data sciences company, and what is really unleashed at full potential here, not only at Biofourmis, but I’m beginning to see it across the industry, is, one way to really think about it is, if you look at clinical practice, medical practice, starting from 2,400 years ago, when Hippocrates basically realize that illness and human maladies are probably not caused by some superpowers and the gods and so on and so forth, but rather, they are caused by something else where it is possible to observe, draw conclusions, and come up with treatment plans. Just the very idea that human maladies was not being caused by the gods, but caused by something else that could actually be treatable. That is what set off the entire era of modern medicine, 2,400 years ago.

But therein lies the catch. The way it is is, you first observe, then you draw conclusions, and then you come up with a treatment plan. So translated, what happens is, all clinical practice effectively starts when the patient visits the doctor, and then the doctor does their measurements and draws inferences and then comes up with the treatment plan and off we go. What technology today has really enabled is something which flips the whole model on its head. Now, because of the proliferation of very inexpensive and very high grade sensors, everything from optical sensors to ECG sensors to even fancier things like galvanic skin response sensors, cameras, what have you, all of these sensors have suddenly become A, very inexpensive and B, extremely high quality. It is now possible to measure human physiology in a continuous manner.

And what has happened here at Biofourmis is that as we look at all these common physiological phenomena, your pulse rate, your ECG wave form, on a continuous way, you can start doing math on these continuous wave forms. You can take the first derivative, you can take the second derivative, you can do correlations with each other and so on so forth. And as you start doing that, you come up with what we call biomarkers. So these are essentially mathematical constructs that seem to be correlated, mathematically correlated with certain disease conditions, certain disease progressions and so on. In many cases, they don’t even have a name, for obvious reasons. It has not been possible to do this kind of measurement and to do this kind of calculation before. But now it is. From a math perspective, it’s relatively straightforward stuff.

As an example, if you took a pulse wave, a wave form, and you took the first derivative, well that’s just measuring the acceleration of the pulse wave. In some ways it’s a proxy for how the heart is beating, how much pressure it’s generating. So even though the first derivative of pulse wave is a clinically understood quantity, you can see that intuition would lead you to believe that the first derivative of a continuous pulse wave form is probably correlated to blood pressure, which is correlated to lots of disease conditions. So you can take these kinds of biomarkers and go off with that. Or alternatively, you can think about, let’s say, voice. We all understand that when you start falling sick, oftentimes, your voice changes. So once again, you can take voice samples of people, of patients, and do signal processing on them, and understand the deviations of their signal processed voice from a baseline when they were healthier, to now when they may not be, and you can again correlate that to disease progression.

So the act of sensing and the act of this massive computing power available, has allowed us to define a whole bunch of biomarkers, which seem to be correlated with disease progressions and disease conditions, which can then be used to provide a signal for earlier intervention, and so on and so forth. So it’s been a fascinating, fascinating new world out there.

Nick Earle:
It totally is. And we’ve had a few podcast guests in the classical healthcare field who are also Eseye customers, and we’ll get onto that. But there is a common theme in that a lot of people are saying is, the old model that you described it, the serial supply chain of, something takes hold, after two or three days, I realize I’m sick, after another two or three days, maybe I’ll ask for a doctor’s appointment after another, I don’t know what it’s like in Massachusetts, but in the London area, after another five days, maybe I’ll get an appointment. By the time I get there, I’m pretty sick, and the … And it’s a reactive model. But what you’re describing is a proactive model, but with the potential to be a preemptive model.

Milan Shah:
Yes.

Nick Earle:
There’s something going on, you don’t know it yet.

Milan Shah:
Correct.

Nick Earle:
But the AI thinks there’s an X% correlation between six factors, therefore intervention, behavior change-

Milan Shah:
Yes, indeed.

Nick Earle:
And I guess you could almost, it’s massively long tail personalization as well, because maybe the biomarkers for one person means something different to the biomarkers for another person, and there’s just no way you can do that with a linear supply chain, reactive system. It has to be personalized, it has to be massive compute and AI/ML, and it becomes a proactive and preemptive, and it’s the holy grail, if you like, of healthcare.

Milan Shah:
Absolutely. I think you nailed a lot of important topics that I’ll tease out a couple. One is the whole idea of personalization. So you’re absolutely right. So now once people have realized the power of AI/ML and how it can be applied to healthcare predictions and so on, the next hurdle that people have run into, so to speak, is the idea of well, AI/ML, how does it traditionally work? You collect a whole bunch of sample data from essentially a cohort group, you train some kind of a model, and then you supply it with data from a new person that it has never seen before, and then it’ll have predictive powers based on that initial training data. Unfortunately, as you correctly identified, every human being is effectively different from every other one. You have your own medical history, you are slated for a knee replacement surgery three months from now, and you just got COVID while you were a heart failure patient, et cetera, et cetera. It’s very difficult to try and find a suitable cohort group that we could train a model on, which will work for you.

So Biofourmis has taken a very different approach. We don’t train our models on cohort group data. We train models based on your own data. So by now, understanding human physiology as well as this basic change in strategy, we can baseline our algorithms usually in three to six hours. What we do use the cohort group data for is to project, if your initial six hour data looks like this, then what will be the effect of nocturnal cycles? What will be the effect of a circadian rhythm, et cetera? We can project without actually having to measure. And so we can start doing predictive analysis between three and six hours from when you onboard onto our system. And after that, it’s completely personalized for you, then the predictions are very, very accurate from that perspective.

Incidentally, a lot of this is FDA cleared, so I love that about the space we are in. There’s a framework in which to measure these claims. So some of our algorithms are cleared by the FDA, which means that the exact, how do you interpret this data and what kind of clinical action you can take based on this data, is documented in a very public form, and validated by obviously a extremely accomplished body like the FDA.

Nick Earle:
Yes. Yeah. Which is hugely important because of the implications of the initial interpretation that you’re making of the data before it goes to the clinician. We get back to approvals in the market in a little bit later on, but I was reading some of the press releases when you did these big funding rounds. Well, actually, what journalists wrote, and one of the phrases that jumped out at me is this fact that you, once you’ve created this knowledge through the AI/ML layer, there’s actually two very different audiences for it.

It talked about the healthcare providers and the clinicians who obviously want to get this data because of the proactive-preemptive bit. And also talked about pharmaceutical companies and the sort of drug approval process, which I’d never thought of. Maybe we can deal with each of those a one after the other, because the second one takes it to a whole new level, to do with how drugs are introduced in the future, which is really interesting. But let’s deal with the first one. So let’s just say I’m a clinician, or I’m a healthcare provider. I would be a potential customer then at Biofourmis, because I want to get this data right?

Milan Shah:
Correct? Yeah, you want to get the advanced warning system, is the more important part. It’s careful because clinicians are already inundated with so much information. The last thing they want is essentially data from a continuous monitoring system that is, forget about a clinician, but humans cannot digest continuous stream of data anyway. So it’s not so much the data that they’re looking for, but the AI/ML interpretation of the data, and the timely signal to take a look at the patient for a potential intervention. And then the clinically established sensitivities as they call it, or accuracies of this signal to know that, “Hey, when this signal is generated, there’s a 95% chance that there is going to be an intervention necessary. So it is worth my time to now take a closer look at the patient and figure out what’s going wrong with this patient.”

So lots and lots of applications across, again, as I mentioned, our technology is personalized, it is also disease agnostic. So we are after, our AI/ML, we will tell you about physiological deterioration. It is interesting to a clinician because then they can apply their skills to figure out what might have gone wrong. The technical trade off of course is that, while we are disease agnostic, well, we are disease agnostic, which means I can’t actually, when I tell you something might be going wrong, actually, my system has actually no idea what it could be.

Where this is applied is in a couple of very interesting cases. So if you take let’s say chronic disease on the patient journey, so let’s say heart failure is a good example. So you’re just turned 50, you visit, you do your annual checkup, and just the fact that you turn 50, you have no symptoms, there’s nothing wrong with you, but you might have a family history of heart disease. So the doctor will tell you, “Well, just because of that, you are at increased risk of heart failure disease.” The typical prescription at that time or advice would be to change lifestyle, change dietary, just because you happen to have a family history and you are at higher risk, right?

Nick Earle:
Yeah.

Milan Shah:
And then you fast forward in what happens, you might start experiencing some symptoms, shortness of breath, something along those lines, and they might put you on statin, so maybe your cholesterol is too high and can’t be controlled, and you could benefit from that. So you could turn around and they might prescribe some statins for you. At some point, that might escalate unfortunately, and you might end up in a adverse event of some kind, or myocardial infarction, or you are completely out of breath and you rush to the ER type of thing and so on. And now you come along with the regimen of medication and so on. And then the doctor takes care of it and then you are monitored closely for X amount of time, and then you’re back into your stable period of management and so on. And you could go from any one of these stages to any one back. You could be perfectly stable and suddenly something happens unfortunately and there’s an MI event, etc.

The value of our system is that along all of these cases, you will typically get an earlier warning, so that the doctor can, especially when you think about some of the acute phases or let’s say you just were discharged from the hospital after an acute event, you are very likely to, I think there’s a statistic out there that says 25% of people who are discharged after an AE of some kind in the heart failure pace, are readmitted back to the hospital within the first 30 days. And the high level reason for that, Nick, is that typically, something like heart failure has to be treated with a cocktail of medicines. We’re also familiar with that. They essentially treat one thing, but that has a side effect, so they have to negate the side effect with another medicine. And so there’s a cocktail of medicines.

Unfortunately, that cocktail of medicines has to be tuned. The dosages have to be tuned for your particular case. And unfortunately, before some of the tuning can happen, you might suffer a second incident and it’s actually the second incident that often is the more deadlier one, has to be treated with much more care, because you’re already recovering from the first one, which is significantly costlier and so on and so forth. So with a technology like ours, what happens is, as you’re going through that tuning system, we will give an earlier indication to the clinician saying, “You need to readjust the dosages, because this person is actually deteriorating.” And that allows the person, the clinician, to adjust the dosages before the second incident actually happen. So if you have been diagnosed, and you are on a dosage medication or a dosage plan, our system will often give an earlier warning to the clinician, to adjust the dosages, which then dramatically improves outcomes and patient quality and so on and so forth. So that’s the typical application, it’s just earlier warning, earlier intervention, therefore smaller intervention.

Nick Earle:
And all enabled by continuous monitoring as opposed to-

Milan Shah:
Correct.

Nick Earle:
… as I called it, the supply chain, as it works. And so that in itself is a huge benefit. But then when, you told me previously about the pharmaceutical companies and how I believe they even came to you and said, “Well, actually, there’s something else we could do with this.” Then I was thinking, “Wow.” It never occurred to me, but then I thought, “Oh, my word, you could potentially reinvent how clinical trials are done here.” So maybe you could share that story with me.

Milan Shah:
Absolutely. So actually, it came as a little bit of a surprise to us as well, when pharma … So what really happened is, we are working on this technology, our focus was cardiac space with the secondary focus on oncology, because that’s where a lot of symptoms show up and earlier intervention can have disproportionately large benefits. So that was our focus, and we are doing our thing and so on, and guess what happens? COVID happens. As it turns out, as I mentioned, our technology is disease agnostic. Long story over there, but our technology got applied to COVID management as well, actually worked very successfully, and really put us on the radar of a lot of practitioners and so on. So we are on this journey, and out of the blue, the big pharma start coming to us. And we are like, “No, no, I’m not really sure why you’re … We’ll gladly take your meetings, but I don’t think our technology really makes any sense for big pharma. It is really about healthcare providers and so on and so forth.”
And they came back and they said, “No, no, quite the contrary.” What’s really happening is a lot of drugs, especially the class 2, class 3 and severe disease cases and so on and so forth, they often come with a library of side effects. If you look at chemo, the fundamental idea of chemo is it’s a poison. Obviously, it’s going to have lots of side effects. So the question then becomes, how do you make these drugs safe? Well, today they approach it essentially as an open loop system from a technical perspective. They try to adjust the chemistry so that the ratio of benefit to risk and so on is appropriate enough for an approval. And then there’s a huge clinical practice that is trying to monitor you and so on and so forth.

So their big insight was, “Hey, what if we use a system like yours to monitor for this side effects? And not only that, but we could even go one step further, and in some cases, pain medication is an excellent example, if you could give us an objective measure of pain, then your algorithms and your system could actually dictate the appropriate dosages for our medication.” So they now have identified a class of a whole bunch of molecules, which are in various degrees of clinical trials right now, where really the only path to an FDA clearance is if it could be accompanied by a system like ours that actually makes the molecule safe to use. Because the molecule will do its thing, we will detect the side effects or we will have earlier detection when something is going off of the safety margins, and that’ll be a cue to the clinician to go and adjust the dosages, or maybe eliminate that particular drug until the side effect subsides and so on and so forth.

And in many cases, that’s really the only regulatory path that is possible. Outside of that, the drug would not be safe to use, and is unlikely to meet regulatory clearances. And if you now extend it to even drugs that are already in the market, Entresto and people like that and so on and so forth, you can imagine that you can see one day where it’ll be almost unnatural to think of a drug that is taken with no control around it. So in the future, we do anticipate a situation where every drug that’s prescribed, even if it’s fairly innocent, it’s just cheap enough to have our system monitor you for any expected side effects, so that we can prevent any adverse events due to the drug.

Nick Earle:
And maybe we won’t then have the day where I get my medication, I open the box, and I find that very tightly folded bit of paper, it looks like it’s been folded by an origami expert, and I opened it up, there’s like eight pages of possible side effects. And one of the reasons was, I guess because of exactly what you’re saying. They give it to a control group, they can’t intervene, there’s no continuous monitoring. So if anything develops anything, they have to record it obviously. But they can’t head it off in the past. They can’t say, “Oh, let’s adjust to it and whatever,” because they don’t have this system. Therefore, the net result down the line is you end up with pages and pages of possible, one patient in a hundred, one patient in a thousand, and all of this. And you read this stuff and think, “Oh God, do I want to be taking this?”

And I guess it’s all because of the way the tests work. I mean, you’re describing a vision where the whole FDA approval of new drugs could actually involve or require, almost, continuous monitoring to improve the dose.

Milan Shah:
Yes, absolutely. Yeah, yeah. Not almost. Absolutely. So what would be regulated, what would be cleared from regulatory perspective is what we call a combination therapy where you take this medicine and you do this monitoring, and you adjust the medicine based on what the monitor is telling you to do. And that combination is what will be approved, or has a chance of regulatory approval. And just the one or the other is unlikely to meet the regulatory safety standards.

Nick Earle:
It’s such a huge subject. I know we could go on for hours, but there are two other areas I wanted to go into if we can. The first one is, you’re a CTO, we are a IoT company, we haven’t talked anything about how you do this. We’ve said AI/ML, but actually, there is a device involved, and that’s where our relationship has been. And I know you initially started off with Bluetooth in gen one of your devices, and then you said, “No, no, no, we really need cellular and ubiquitous, constant connectivity everywhere,” which is where we came in. Maybe you could talk a little bit about that.

Milan Shah:
Absolutely. So especially driven by COVID, there was a realization, plus all these use cases that we talk about, whether it’s the pharma side and drugs, new drugs coming out the market, or just the number of diseases that we could manage and help improve the management of. You can imagine that our initial solution, which included obviously the data science, but data science needs input in order to drive it. That input as you mentioned, comes from a wearable. So ours happens to be an upper-arm-based wearable, has a plethora of sensors in it, which produces the signals which are then fed and digested by the AI/ML and so on and so forth.

Our initial version, the current version that’s in the market is a CE-approved device and connects over Bluetooth to a smartphone, and the smartphone is the one that actually transfers the data to the cloud. The system works, people manage COVID with it, for example, in many countries, actually, at a nation state level and so on. But what that has really done is just driven the realization of the value of this into the future, the potential that something as a solution like this can bring to bear. And a few countries, so because of COVID, a few countries actually chose our solution as the way to manage COVID in their country. So if you got COVID in let’s say Australia, the standard protocol will actually send you home with our solution, and then our AI will indicate when you need to be brought back in, perhaps for a ventilation. But with the vast majority of course recover without any ancillary admissions and so on and so forth.

But people like that, countries that, have come to us and say, “Hey, outside of COVID, we could actually use this for many other disease areas.” The problem of course is when you’re trying to use it at a population-level scale. Trying to create an ecosystem where there is a controlled device, which is a smartphone, which obviously has to meet configurations and uptime numbers and so on so forth, that just is not technically viable today. Nobody’s going to carry two smartphones to begin with. And even after that, trying to keep the device in close proximity to the Bluetooth smartphone has actually turned out to be extremely challenging. People walk into the bathroom, people walk into the basement, they go to the kitchen, they leave the phone charging in the study, and they have gone to sleep in the bedroom and now Bluetooth connectivity has been lost.

So Bluetooth is actually just not a viable way to do this continuous monitoring. And that’s what led us to saying, “Hey, maybe we should just put a 4G chip into the device, have it connect directly over the 4G network into the cloud,” and now the patient just has to slip in our device and they’re in contiguous monitoring, and magic starts happening. So that was the promise. At this point, that is the reality as well, and that is really what introduced us to Eseye. So we looked for a way to make this wearable work across the world in all of the geographies where we already have business, and then more, and ultimately identified Eseye as the partner. And even though we didn’t realize it at the time, what’s turned out is just the whole getting an IoT device to actually work in an approved manner, pass all the regulatory hurdles in across the world, is actually a very, very tall task.

Nick Earle:
It is.

Milan Shah:
And certainly, obviously, it’s engineering and science, so it’s doable. But you can imagine for a company like us, where we have so much potential just in our domain, for me to try and allocate engineering resources to really become an expert in the 4G network part, is probably not the most direct investment that we really would’ve liked to make. So a partner like Eseye has had disproportionate amount of effect. And even though we stepped into it more on reputation and recommendation and so on, what we actually experienced was Eseye’s tremendous experience and expertise, in not just in what they do, but in extending their expertise into what we do.

So today, engineers and support people at Eseye actually have inordinate amount of expertise on my device and what it’s trying to do and the implications of how it does things and so on. And that was necessary. That overlap of expertise was what was necessary to bring what we now have to the market. And that has been a big success story from our side. So today, our next-generation device is available and works across the world. I believe there are three countries which I didn’t even know they were actually officially recognized as countries, but any case, except for those three countries, the device was-

Nick Earle:
Yeah, that’s right. People ask us, we say we have global connectivity and we solve this problem which, by the way, everybody thinks isn’t a problem, because as we said in many podcasts, I just put a SIM in my phone, it works, doesn’t it? Well, IoT, no it doesn’t. IoT devices don’t work like that. But every customer, it’s an education process. But yeah, there are three countries, and it’s more to do with the political regime-

Milan Shah:
Yes.

Nick Earle:
… and what those guys do with technology. So there are three, but other than that, it’s every country in the world. And yeah, no, it’s been a great journey, and we’ve learned a lot together. And trying to do this in a relatively small device. And as you say, not only can the patients perhaps leave the phone downstairs and go to sleep upstairs, but some of these patients are in their seventies or their 80s, and telling them to sync with a smartphone, I mean, it’s just not going to happen. So cellular around the world, with the one-button press that we’ve talked about, and getting this device, because the device is the … It all starts with the device, and then the data goes to … And getting the data into the cloud, another technical problem. People think it’s easy, it’s not. From a variety of different operators, and then getting into your AI engine. And so it has been a great journey.

Let’s try and bring it all together if we can, Milan. Such a great story, but let’s try and bring it all together with one example of one country. And I know that, I think you’ve mentioned Singapore earlier. I know that you guys are pretty active in Singapore. I think there’s what, there’s 7 million people there or whatever, beautiful place. Maybe you could just describe a little bit about what you’ve been doing in Singapore. Because I believe, you talked about regulatory approval, and I believe you’re pretty far down the road there.

Milan Shah:
Yeah. Yeah, yeah. Thank you. So Singapore is somewhat special in our history because we were founded in Singapore. Our two co-founders happened to be students at the National University of Singapore, pursuing their PhDs, when they decided to do a Biofourmis. And the country of Singapore has been a tremendous aid all along. They’re actually an investor at this point. EDBI is an investor, early investor, I should say. So they were very much a important reason why we were able to get to this stage of our journey. But interestingly enough, what really happened is, again, COVID happened, now Singapore, Hong Kong, all these regions, they had a previous experience with SARS. So they had a tremendous experience in how to manage coronavirus like COVID prior to that. So they weren’t necessarily looking for technology like ours. And in fact, in the grand scheme of things, they were very successful at very aggressively managing the spread of COVID.

And then of course what happened is, eventually an outbreak did happen. In Singapore, there are these higher concentration living arrangements, they call them foreign worker dorms. And COVID broke out over there, and suddenly, they had, essentially overnight, or in very short duration, thousands and thousands of COVID patients, all in a very concentrated physical location type of thing. And what they decided to do was basically take every resident of these dorms, give them one of our solutions, and then they literally set up a clinical center right at the basement or one of the floors of the dorm, where they brought in some clinical staff over there who were monitoring all of the patients, all of the subjects, and then calling the ones where we signaled would need attention, calling them into the clinical center that was set up in the dorm itself.

So that’s how they managed COVID outbreak, when it did happen in these highly concentrated areas. And that is really what led them to think about, “Wait a minute, we see how this works. People are going about their lives, they’re wearing this device, they’re being monitored, when it signals, that’s when they’re brought into the hospital.” And they started exploring with us the idea that, “Outside of COVID, can we use this for many other disease areas as well?” And the answer of course is yes.

I mean, what really happens is exactly like you described, if you yourself fall sick, you’ll wait for three days until it hits a certain threshold, and then you’ll seek an appointment and then you’ll finally get a visit with the doctor. By that time, A, the disease has progressed, and worse is 90% of the time, there’s not much to do. They’ll tell you to sleep it off, if you haven’t already done that already. So this whole process is inefficient on both ends of the spectrum. A, it doesn’t get to the people who need it early enough, and then it wastes way too much energy looking at patients that actually don’t need-

Nick Earle:
And if I can, and this is really the nub of it, it wastes way too much energy and way too much money, because what you’re describing is, you’re describing that the customer is the government. The traditional RPM, remote patient monitoring and healthcare companies, are selling to the clinician. Persuade the clinician. But you are talking about an environment where you say to the government, “Look, if you go from reactive to proactive, preemptive or continuous monitoring, you’re actually going to save potentially hundreds of millions, maybe billions of dollars in your national healthcare infrastructure.” A lot of them aren’t going to go … In the first place, they’re going to have an intervention. So actually, you’re selling at the government level-

Milan Shah:
That’s right.

Nick Earle:
… a medical device for the population.

Milan Shah:
That’s right. That’s right. That’s exactly correct. And it’s just a safety mechanism, exactly to your point, for those countries that have nationalized health systems. It is a way to achieve a tremendously higher efficiency in terms of allocation of resources. You are the NHS, you are the Ministry of Health in Singapore, and you’re trying to figure out which patients to allocate your resources to, something like this can be essentially disruptive and game changing, because we are bringing continuous monitoring to the table at this point in time.

Nick Earle:
And wow, tremendous. And that model, when you think about the potential globally for that model, it’s truly disruptive in a very good way. And as you said, it’s not disease specific. I think you used to phrase early on I quite like that the AI/ML doesn’t know. It’s data and interpretation. It could be heart condition, but it could go across so many different areas. And you guys are one of the first movers, you’ve raised the most money, and so clearly, you’re pushing hard, and I know working with you on almost daily basis, well, we are working with you on a daily basis, you guys are running hard, charging hard in many different areas. And it’s such an exciting IoT case study, which is what these podcasts are all about, and-

Milan Shah:
Absolutely.

Nick Earle:
I think it’s a tremendous story. Let’s finish if we can, we could go on for hours on this, but let’s finish if we can. Is there a vision? It’s so big. You think how you could do this, you could do that, you could do this. Is there some vision for how it just all becomes part of our daily lives?

Milan Shah:
Yes, Nick. Recently, I was talking to someone and an analogy came to mind, which is, if you think about just 15, 20 years ago maybe, GPS was a new thing and you went off and bought this $2,000 piece of equipment that sat on a big sack on your car dashboard type of thing. And GPS has now become completely ubiquitous. Either your car has it or certainly your phone has it. And today, while of course it’s entirely possible to transport yourself in cars and automobiles without a GPS, practically nobody does it. Everybody will, even on your daily commute, you want to take a look at the best route, given all the traffic.

Nick Earle:
And it’s dynamic. It’s more intelligent than its…

Milan Shah:
Absolutely.

Nick Earle:
… itself. Yeah, yeah.

Milan Shah:
Exactly. And it’s really no cost, it’s right there, it’s something that has a tremendous benefit and really not much downside to it at all. And everybody uses it. It’s almost impossible to think about landing or trying to go somewhere without a GPS these days. Well, the same analogy applies to healthcare. Whether you’re navigating a disease or you prescribe a complex molecule or complex regime of a medication, trying to navigate that journey, completely blind with no guide at all, just will seem very, very antiquated in a very short order. Systems like ours are likely to become very ubiquitous. People will just have it, whether it was because the government of Singapore just gave you a device and it’s just easy enough to wear it and now you’re navigating your health without really having to do much about it, just wear a device and the system will tell you when you might need to get some care, or as we described, you’re in some complex medication regimen, then you have a system like ours that is guiding you through that journey as a patient.

I think systems like ours are going to become completely ubiquitous. It’ll be impossible to think about the day and age when we were trying to navigate personal health with absolutely no guidance like you’re saying on the basis of a little sheet of paper that was folded together by an origami expert. I mean, that’s the equivalent of those-

Nick Earle:
Which nobody reads. Well, very few do. I’ve read a few of those, but I’m not going to read them again. They scare you. But they do say that technology truly becomes ubiquitous when it’s, in effect, invisible.

Milan Shah:
Yes.

Nick Earle:
And what you are describing, it’ll be, you notice it more by its absence than by its presence.

Milan Shah:
That’s correct.

Nick Earle:
And that’s exactly what you are describing. So Milan, I think we’re going to have to leave it there. It is such an exciting story on multiple fronts. I mean, it brings all the pieces of IoT together, but it also addresses such a big problem with such a huge potential opportunity. I mean, selling to governments, like you said, the government funds the device, so it’s a different commercial model, something that’s transposable across different human conditions. The pharmaceuticals and the healthcare providers being different groups. It’s a really exciting thing and we’ve been very happy with our partnership.

As you say, everybody always thinks the device is easy, and every single one of these podcasts is clustered around that central problem. I thought the device was easy, and then I found out it wasn’t, so came to Eseye. So thanks for that. I know there’s lots of exciting things in the future, which we can’t talk about, but in the meantime, it’s been a really, really great podcast. Thanks for joining me, and I’m sure our listeners loved listened to it. So thanks for being my guest on the IoT Leaders Podcast.

Milan Shah:
And Nick, thank you for inviting us. Really appreciate the time.

Resources

Whitepapers

2023 State of IoT Adoption Report

Learn more

Blogs

Connected Health: the True Value of Good IoT Connectivity

Learn more

News

Eseye Named the #1 Champion eSIM Connectivity Provider

Learn more

Blogs

Core Challenges that All Industries Face with IoT

Learn more