Epitel - From Research to $26 Million Series B

Mark Lehmkuhle - Epitel

Season two, we did forty two different episodes of founders and product leaders all across the country. in season three, I think we're already up to like sixty, sixty one, which is shenanigans. It's insane. Wild to believe that. but today we have a super exciting couple of guests with us from EBITL, Mark and Jay. Welcome to the show. Thank you for having us.

J Carroll 01:16

Yep, likewise, thank you. DeAndre + Grant (01:16.763) Awesome guys. Yeah, to get this kicked off, you guys have exciting news and working on incredible technology and medical device world. would love to get it kicked off on just kind of quick intros on yourselves and and a little bit about the company as well. And I'll pick so you don't you don't have to guess who's going first. Jay, let's go let's start with you on background and and what we're here for.

Yeah, absolutely. I'm I'm the director of software engineering and architecture here at Epitel. And my background is extremely eclectic. I actually it I I won't go through the full laundry list, but it began in childhood education like I don't know, like fifteen, twenty years ago, thinking that's what I was gonna go into. And somehow I made my way through data science and eventually made my way into medical device engineering. it's been quite the journey.

these days, I'd say over the last ten to thirteen, fourteen years, I really specialized in blending medical streaming telemetry or streaming technologies in general, along with a lot the the heavy governance that comes working with an agency like the FDA. and as you can imagine, you work with PHI and you work with streaming. you're trying to marry the two worlds of hardware and software and an ever moving target of technology. And it's kind of where I found my niche and a lot of what I do here at Epitel.

DeAndre + Grant (02:42.655) Amazing. And then Mark, what what role do you play in in all the the craziness that you guys get to do every day?

Mark Lehmkuhle - Epitel 02:47

yes. I'm Mark Lamquil, the CTO and founder of Epitel. And my background is in biomedical engineering. Historically, I've been in academia, so this is my first company. So Jay has way more experience than I do in the medical device space. But yeah, so this is kind of my brain child, I guess you could say. DeAndre + Grant (03:09.457) And it it escaped the lab. And it fully escaped the lab and here we are. I mean this this is the fun part about the really technically driven devices, is they have to start out in a lab that was doing something that didn't exist. Right. Like this is like everything that we take for granted today of modern medicine was someone's crazy idea that accidentally went too far and broke out of the lab.

J Carroll 03:09

Trial by fire.

Mark Lehmkuhle - Epitel 03:11

It has. It's still running around here somewhere. DeAndre + Grant (03:35.037) Yeah. So so Mark, what what are you guys building today and and what what do you guys do for all the listeners aren't familiar with Epitel? Mm-hmm. So we, I would describe us as we developed the machine learning for the detection of neurological disorders. And we're starting off with seizures and we just so happen to have our very own wireless wearable EEG sensor. It looks like a little chicklet. It has two gold electrodes on the bottom, attaches to the scalp with.

these one piece stickers that's both an adhesive and a conductive hydrogel. So there's no goopy gels or anything like that. And the reason why I say we're a machine learning company is that we create lots and lots of data. And what we're trying to do is really convert that data into clinically actionable information. And the sensors themselves are simply a means to the end, if you will. DeAndre + Grant (04:24.261) you mean the necessary evil of hardware? Yep. This is this is the whole world in which we all live in and and I think it's the the exciting time and Deanna, you and have talked this on a bunch of episodes this season is there's this tsunami of a theme in both medical device for sure, but even in larger IoT, transitioning from remote what I'm gonna call binary control or binary monitoring of is the thing on or off or what is the temperature, right? Of that has been most of IoT for last 15 years. Locked or unlocked, you know, what's the state of this light bulb and

Yes, exactly. DeAndre + Grant (04:53.809) All of a sudden, with machine learning getting really good in a bunch of creating creative and fast in your ways, we can take huge amounts of data that is noise. It's effectively noise. And you can take enough data with enough measured outcomes somewhere else outside of the device, put them into software magic. And what comes out the other side is a prediction of we think this thing is currently happening with some real amount of certainty, which is what's really neat. And that's enabling everything in med device and other IoT things like energy monitoring or leak detection in your house.

But it is the hardware is just the necessary evil to collect this big old stream of data that the machine learning makes actionable. That we had this data years ago. We just didn't know what to do with it. Mm-hmm, yeah, absolutely. just, you know, there really hasn't been any way to get this data for long periods of time. know, EEG's been around since the 1800s, but not in a long-term usable format.

J Carroll 05:51

Yeah, I think DeAndre + Grant (05:51.328) Right. It's one of those like you you had the little ink on the paper, right? Little arm that was gonna move and give you the single. And you could store that and read it like a human, but you couldn't process it. And then we take digital ones, but it was just in the hospital, not on you all the time. So you had a snapshot in time, not a huge amount of data. And Mark, what w one of the first questions I've got for both you and Jay is just around, you know, you guys just closed a twenty six million dollar series B. so congratulations there. But let's rewind back a little bit, right? You guys this didn't start off originally from the technology perspective, Mark.

Okay.

Mark Lehmkuhle - Epitel 06:06

That's correct. Thank you. DeAndre + Grant (06:20.799) For humans, right? You actually started somewhere else. We'd love to kind of pick apart that story as to how you've, you know, we we understand who you where you are today, but kind of give us a little bit of a background on where you got started. Yeah, absolutely. you know, again, my background's in biomedical engineering and academia, and I was at the University of Utah. And the University of Utah is unique in for the past nearly 40 years now, they've had what used to be called the Anti-Convulsant Screening Program. And so this was a service through the National Institutes of Health, where if I'm a pharmaceutical company that might have a novel anti-seizure compound,

I could effectively send it to the NIH, the NIH codes it, sends it to Utah, and then it gets run through a battery of tests. And then promising compounds would then go on to clinical trials. And it was a way to de-risk the development of these new compounds because there's so much noise in these clinical trials. And what that noise does is it just creates a very expensive trial. So it was kind of preventing pharmaceutical companies from wanting to even develop new.

central nervous system drugs. And so we were kind of towards the end of this test battery where we were testing the most promising compounds and mostly rodent models of seizures on and off the drugs for almost the lifetime of the animal. And these animal models, much like humans, may only have a seizure once a week or even less. And so the only way to really capture those events was to record EEG 24-7 for nearly the lifetime of the animal. And so, you know, nothing out there existed. So we effectively built our own. was pretty slick concept, you know, kind of based on the most similar would be the wireless tire pressure monitor in your car.

J Carroll 07:52

Okay. DeAndre + Grant (07:58.985) Forever, yeah.

Mark Lehmkuhle - Epitel 08:14

And so we didn't, it didn't actually transmit anything. It created an electric field and we encoded the EEG into the electric field and then decoded it through a platform that was underneath, you know, underneath the animal. DeAndre + Grant (08:24.883) those stationary, right? So you could do a very low power transmit of the signals, just you know, pick up the E G, turn it into a different frequency or or radio wave, read that somewhere else and then encode it. Right. And like basically put that into memory.

Exactly. And with these thumbnail-sized transmitters, we could record for six continuous months.

J Carroll 08:35

Yeah yeah. Yeah. And and and I wanted to add a a very quick anecdotal story and I think I've told this to Mark before, but I I did a stint in academics for about seven years before going into medical device product development and stuff. And w part of what I did for a number of years was w as a graduate student was I worked in a rat research lab and I cannot underscore how thankful DeAndre + Grant (08:43.454) That's awesome.

and frustrated I am that Mark developed technology when he did, because when I was leaving my rat research lab, I was the graduate student, a part of an army of graduate students that used to go in on a daily basis. And we would have all of our genetically selected or or bred rats, and they would have a head stage on them with a cannula with electrodes that went through it. And my job was every day, I think it was like six AM for an hour, I had to go in there.

Mark Lehmkuhle - Epitel 09:16

you

J Carroll 09:35

and adjust the head stage, wire it up with a wire, put them in their maze, measure them as they ran around, and sit there with a little notebook watching an oscilloscope, waiting for a click, a little audible click on on the oscilloscope. and then do like a manual FFT, you know, w if we found a data my the instruction the instruction DeAndre + Grant (09:56.199) Like, yeah. Hit freeze in the oscilloscope, run the FFT.

That's exactly what yeah, exactly what it was. And and the instruction all the graduate students, there's an army of us that had was it doesn't matter where the professor was, because it was implanted in the occipital lobe and they were doing, you know, s it was a NIH grant, I think that we were working on. It was so hard to find these neurons firing and all the stuff that we were doing. And if if we had any problems or if we ever had a click on the oscilloscope, a click of any kind, go and hunt him down on campus and immediately bring him to the lab.

And that was that was hell for about two weeks. Dr. Calton, if you're watching, I'm really sorry, but that was a really difficult program for me to participate in. But we got through it and I wish I had Mark's technology back in the day. Would it have helped that project? Probably not, but it's still very inspiring to see the problems that Mark has been able to solve with his technology and how Epitel's been able to grow into it.

Mark Lehmkuhle - Epitel 10:32

you DeAndre + Grant (10:52.839) Well, and I think the the magic of what you just uncovered is this is where academia does things that product development can't. Right? Product development in like engineering for to bring a product to market can't afford to take too much risk. We can't go to do things that are too left field or too in too I'll call insane. you know, we have to do things that have a certainty of working on the timeline and we'll say, we can't do that. Or, yes, I can do it and here's I'm going to do it, and here's about how long. But in academia, you're like, Well, someone's got to do it. We have to figure this out. So we're gonna go

throw stuff against the wall until it works. Mark, how many iterations of how do I get data off the mouse without a wire stuck to it did it take to get to what you ended up like what works the first time? Because I'm sure that's not what works today, but that was what worked the first time. Yeah, Great question. I mean, it took a lot of iterations because, you know, once you get one electrode, now everyone wants two electrodes and then four electrodes and six electrodes. And, let's make it smaller for my neonatal model of seizures. So you can imagine that the project snowballed from there. And so we did make a few different iterations of this and we're selling it to researchers all over the world.

DeAndre + Grant (11:45.085) How hard could it be? doing these preclinical model developments. DeAndre + Grant (12:07.559) Mark, I want to dive a little bit more into the the electrode side of things. But before we we go into that, Jay, whenever we first met, you you and you kind of mentioned it briefly at at on your introduction. But I think it could be cool to kind of talk about your story a little just really quick briefly on if folks specifically for people that like want to break into the world of hardware, product development, medical device, et cetera. if I remember correctly, you took a pretty big leap. You know, maybe you lived on the West Coast, you came to like the East Coast kind of thing, you kind of had a cool story there. So kind of wanted to kind of feature you real quick and

Can I p pick that apart a little bit?

J Carroll 12:39

No, I I I appreciate that. It's very flattering. I I think the the the story is probably far more glamorous than than actually living it. I was in graduate school. I I was teaching over at CSUS Sacramento. I was emphasizing a lot of research methods, advanced statistics, working for the state. I was working like three or four jobs. eventually kind of abandon all hope, you know, because it gets kind of exhausting. And I I I decided to try my stint from from Sacramento. It's just a hop, jump, and a skip over to Silicon Valley. And certainly, you know, anybody can make that work, except for me, apparently. So for about four years, I was just, you know, on the grindstone as a data scientist over there and a statistician, just trying to make it work. And and back in the, you know, the

early two thousands, two thousand tens. you know, data science wasn't as glamorous as it is today. Apparently I jumped ship way too early and should have just stayed in the industry. I'd be a millionaire by now, probably. but I had some friends that were leaving data science for programming and it was very lucrative and so I decided to throw my hat in the ring. I got a call from a company I've never heard of before.

asking if I would like to leave Silicon Valley and start work in Boston. And I had about four days to quit my job and get in a car with my last nine hundred dollars and drive out to Boston, which is exactly what I did. I made the entire trip and I think it was in just over twenty four hours. so it was a it was a long haul with my tiny little Silicon Valley car. and I joined a company called Zool Medical. I went from not doing medical device developments and focusing on you know, data science and crunching numbers, report generation, large-scale computing systems, switched gears from there and immediately jumped into the deep end working on class three medical devices. if there was ever a person that suffered from imposter syndrome,

working along other people who have biomedical degrees, PhDs. A lot of my peers that I was working with at the time, they you know, they were born and bred and raised to be engineers. and here I am coming from Silicon Valley and West Coast cul west coast culture into this east coast environment. And hit the ground running. and didn't look back. I think it was actually a beautiful marriage of my data science background and my intimate understanding of research methodologies and statistics that's led me to a lot of these days I do like A B experimentation or try to emphasize that for product development. and I'm

You know, hopefully Mark agrees with this statement, but in a lot of our conversations, I try to be extremely analytical and extremely data driven. hopefully I have paid my my my dues through the gauntlet and can be considered a peer among other purist engineers. but certainly my background is is very, very different, I think, than a lot of folks. and I don't know. My my journey, while I feel it's unique.

I feel like it could be anybody's journey. you just you wait for those opportunities and when the opportunities arise you need to make a decision. Is it something do you take your last nine hundred dollars and travel across country to work for a company you've never heard of, working on class three defibulators and monitors and ventilators? Or do you stay in Silicon Valley s you know, slumming it with the the rest of of the crew out there that's trying to make it work to be

you know, silicon millionaires and I took the jump. And it's worked out really well for me and my family. DeAndre + Grant (16:54.483) That's awesome. My and I love the thought that always goes through my head in those moments is how hard could it be? Right? Yeah. For anyone who's thinking about doing the journey, it's hard. It's not impossible. intellect, determination, opportunity, friendships. there's been a lot of things along the way. I've had a lot of really fantastic mentors and even today working with Mark, and I I I my boss currently is Matt Parrott.

we have a lot of really fantastic mentors and leaders at Epitel that have come together. it's been such an an assemblage of leadership here that's been so inspiring that's made the jump over here very easy for me. And I'd I would advise other people thinking of careers or looking to jump and or blend their backgrounds. there's opportunities and then there's there's people opportunities and sometimes it's the confluence of both necessary for for it to to make it happen.

DeAndre + Grant (17:53.404) Absolutely, Jay. No, I thank you for for sharing that. I just thought it was a super cool story. And I think it's just similar, right? Like with you as well, Mark, just taking the leap and betting on yourself and building something. And super cool to see that you guys are both now now came to Epitel and you guys are building what you are. but I just want to definitely feature that because I thought that was super cool whenever we first had that conversation. so kind of getting back over to the hard tech side. Sorry, that was a soft side. That's the stuff that I love. getting back over to the the the technical side of things. Let's talk about electrodes. So when it comes to like this closely pinning of electrodes.

Things can kinda get fuzzy. It's hard to measure things. Kinda walk us through your process and h on figuring out what was the right way to approach that, especially when it comes to the product today.

Mark Lehmkuhle - Epitel 18:30

Yeah, yeah, great question. So, you know, when I started the company around that time, iRhythm was coming out with their wireless halter monitor, the Zeo patch. And, you know, we kind of thought, that's great. It's replacing the seven lead halter with a single lead patch that could last much, much longer. We could do the same thing for the brain. But, you know, the challenge there is that in a traditional EEG, you've got electrodes that are spaced so far apart. And it's really to be able to pick up those fields.

And so the question was, well, how many electrodes do we actually need? Because we know that, you know, putting on multiple things is going to be difficult for people and we want to make it as easy as possible. But we also want to be able to, you know, detect the signals that we're looking for, in this case, electrographic seizures. And so we kind of recreated a study from the early 70s that started removing electrodes from a wired EEG and asking

can we still detect the electrographic signature of these seizures? And it kind of came down to these four corners, two in the corner of the forehead and then two behind each ear, roughly over the mastoid. And then it came down to, so we've got our four locations, that's the minimum that we could do. We can still detect the electrographic signature of these seizures due to volume conduction. So a neurologist can still see the evolution of the EEG into the seizure. It might not be as strong as what they might be used to with an electrode right over top of the seizure focus, but they could still see it and detect it. And then the question was, you know, how small could we actually get these sensors? Because, you know, the bigger the thing is, the less people are going to wear that as well.

And so, recreated a more different study that started moving electrodes closer and closer together, showing that roughly when you get them about 10 centimeters or 10 millimeters apart, you're recording basically the same thing from the scalp. And so, that kind of was the lower limit. And effectively, we came up with a halving medium for related developing. how close we can get those electrodes and still detect a signal. You know, it does change the signal a little bit, but it's it's still ideal for a neurologist to be able to detect these electrographic signatures.

DeAndre + Grant (20:52.681) When I think the d digging into the physics of what's going on, both on the measurement side and of the how humans interpret data side, before we had machines that could watch this and compare lots and lots of things, and we also knew what we were looking for, we needed a really high resolution, really obvious signal for our brain like, yep, that's a thing going on. And I can see it here or here or here, and we needed all this data. Once you know what you're looking for, and once you can have a computer help you find that signal in the noise, you can deal with way less data.

J Carroll 20:52

Mm mm.

Mark Lehmkuhle - Epitel 20:57

Mm-hmm. DeAndre + Grant (21:21.885) like infinitely crappier data too, not even just less, just crappier data. And the the other thing that I think is really neat, and I kind of relate to the current AI revolution, is there's all these really big heavy frontier models that can get it right on the first shot. And then there's all these plucky little local models that if you let them grind long enough, they'll converge at the right answer. Right. And I think you guys are s going down that same ideal pipeline of, hey,

Yes. Mm-hmm. DeAndre + Grant (21:48.2) You're in a lab and we need to be able to catch an EEG in the moment live. We need a ton of resolution in data. I can watch you for hours because you can just wear this thing around. It's no big deal. I can look for way less crappy data and still see it happen because I see what baseline is. I can track all these differences. Is that kind of how you guys kind of came to this conclusion of our scientific instruments could capture this big high resolution thing, but we don't need as much of it when we know what we're looking for?

Yes, absolutely. And it's also kind of, you know, we're trying to make a long-term EEG recording solution and how do we make that as inexpensive as possible? And so even from a machine learning perspective, you can think of like these very large foundational one-shot models that might be computationally expensive in the cloud. Eventually we want to bring all of these models back to the person and say a personal area network of sorts. You know, how do we make that efficient in the cloud? So

that, you know, Jay can optimize this so we can scale it on a national level and still be inexpensive to deploy.

J Carroll 22:49

Yeah, and that's I yeah I wanted to add to that. I I think, you know, we we a number of times in this conversation we've been talking about sort of like the evolution of thought. And, you know, I I think the very next thing that happens after we accomplish sort of like what Mark is is talking about, then I I I already know as as a technologist, it's gonna be okay, what can you do to move it to the edge? Okay, what can you do to move it to the to to to the local, to the phone, to the device?

DeAndre + Grant (23:12.351) Yeah. And right, and 'cause ultimately it's it's just, you know, it's it's it's pushing that frontier. And I think I wanted to ask Mark a question. I hope it's okay, Mark. Yeah, or or DeAndre and Grant ask as a as a host question here. but I I I'm wondering if, you know, you kind of see the four sensor model or the montage there.

Mark Lehmkuhle - Epitel 23:23

Awesome.

J Carroll 23:37

you know, we we got there number one, because we we kind of challenged the status quo. You challenge the status quo that, you know, for for what we have there is is what certainly what everyone would would need. And that's what's been used for decades. Don't change it. And then we've been able to come in and say, you know what, I I think there is a better way to do it. Let's be bold enough. kind of like I think what Grant was saying earlier, it's academics can't purists working in academia sometimes can't always afford to take that risk.

you took that risk, I think, and that's that's been kind of instrumental or or a prime ingredient. But then the other bit that I think is I I would like to hear your thoughts on is whether or not you know you would expect further evolutions on the the the sensor topology or the sensor layout further given that you know if we're looking for EEG signal for seizure activity maybe four is optimal today, but as we move into other modes of neurological signal detection, maybe we don't need four, maybe we can go down to one, or maybe it requires ten. I mean, do you kind of see that a that moving target as as we step through this journey?

Mark Lehmkuhle - Epitel 24:51

Absolutely. mean, you know, just like the evolution of AI ML models over time, know, the tech stack and our sensors is evolving over time too. know, components are getting smaller, they're getting less expensive. We can do more on smaller components. So moving the data analysis, whether it's digital signal processing or what have you, to the edge of the sensor itself is becoming more more doable for sure.

The other part of this too is, when I started the company, it was really for people who had already been diagnosed with epilepsy. Because we knew that people, you know, going back to clinical trials, we knew that the primary outcome measure for clinical trials is self-reporting. But study after study was showing that self-reporting was maybe 50-50 at best. You know, people don't know when they've had a seizure. A lot of seizures can occur at night while you're sleeping. And then other seizures were going misreported. You know, people thought they had a seizure and...

they didn't actually have a seizure. So, you know, let's create a system that can simply count the number of seizures that you're having. And that concept in my mind only used a single sensor. You know, we already knew where the seizures were coming from. Let's say someone had right temporal lobe epilepsy. We would just need to sensor behind the right ear, let's say. And, you know, that would be inconspicuous, kind of looks like a hearing aid, which is more socially acceptable. We know that putting anything in kind of the facial area is going to be difficult from

social acceptability perspective. And adherence is something that's always on our minds. So, you know, making something that people will wear and want to wear day after day is definitely a real challenge, an engineering challenge. DeAndre + Grant (26:37.023) Well, and it that was the most fascinating full circle for me to to watch. So we started with a question that was insanely technical. Like where do you do the computing? Which like the ping pong from I have to do it on my device because the internet's not good enough to get data to the edge or the cloud. Then everything went cloud. And now we're trying to pull everything back to the edge for some percent. You like the ping pong's great. And then it ended with the most user focused product story I've ever heard of, coming from someone who left academia to be in a product. Well, I don't care how good the tech is or how great the single noise ratio is, none of this matters.

If they don't wear it, what does the human think? That to me was just like the coolest, like full circle of like, Mark, you've lived all the lives. You've you've been a researcher, you've been in academia, you've been in product development, you've been an engineer's engineer. Like that was all those answers in one. you I would say the tech stack is certainly tractable. The really hard part is human factors.

DeAndre + Grant (27:29.053) Mm-hmm. why? Just because, you let's think of forecasting. There's a lot of movement in our area of epilepsy to do seizure forecasting or seizure prediction. But what does it mean? If there's a 40 % chance of rain, you might take an umbrella and I might not. And so there's no universal what is good in that space. And so that's why I human factors is a real challenge because in our population, we go from neonates to someone in their 90s. And that's a huge spread of age and experiences for those folks.

DeAndre + Grant (28:14.311) It's such a unique design challenge. You're exactly right. Well, and the thing for me is like, do you affect your own outcome? Like, if you're a user who knows they're 75% likely to have a seizure that day, do you change your behavior to make that 75% prediction wrong? Right. Like how do you, Mark, calibrate your data set? Because if you start telling people what, you know, the crystal ball, they're gonna change their behavior in a way that may change the outcome of that crystal ball. It's like the self-fulfilling prophecy or not.

That's correct, yes.

J Carroll 28:39

Yeah. And and I think you're hitting on a larger problem too, which is you know, put drawing from data science and and the research arm of my background, you know, we had a lot of problems with garbage in, garbage out, right? I think we've probably all have heard that. And the reality is is when you put these products out into the field and you're not in a clinical lab sterile setting to supervise

the purism of the attachment of those electrodes and the use of the data collection devices and all of the variabilities that occur for data transmission, stemming from the sensor to the phone to the cloud and possibly round tripping some of that information. If you don't account for those things, The reality is is that if you don't have the observability hooked up or if you're not aware of of the the confounds that are impacting your data collection, it how good is your data as it's going into in into the cloud? And ultimately if you're using that data to do any sort of training on, you know, you could just be building on a faulty foundation. So you can't in my in in my book in engineering, you know, for the way that I lead my teams, usability and human factors.

is not a second or a thirst class citizen. It is or third class citizen is a first class citizen. if no one will use our product that we built, or no one can use it reliably and consistently, then we we're not done doing the product development or the engineering portions of it. DeAndre + Grant (30:10.845) And it's it's one of those things that you said it so well. The sterility of the lab is the thing that we all take for granted when we're doing prototyping and testing. And then no plans for each first contact. You get out in the real world and humans are messy and lazy and different and crazy and all, you know, all these adjectives that just mean, you different from each other and different from this perfect environment. As an engineer, how do you design the thing you can afford to design that doesn't make the product worse for most people to include?

Include the most number of users that will use the device. Right? You can put so many sort safeguards in a device's architecture that it will always be used right, but then it might suck to use. Right. It might not be frictionless enough, it might not blend in with your day. And you can do the opposite. You can make something so simple and easy to use that a 20% of your market will never use it right because they are just using it wrong and don't know that it's wrong. And it's the that balance of who do you design it for? And then how do you tell both

In the macro, did I design the right product the right way? And in the micro, can I tell in the moment that this data is good before I ingest it? Right. Jay, you can take a ton of data, even though it's bad. And you don't care if you took bad data if you know it's bad. Right? I can tell the user didn't put the sensor on exactly aligned on the electrodes to the hydrogel. We're gonna get some, you know, bad impedance. Let's just throw out today's data. That's a way less bad scenario than, well, I have no idea out of these thousand patients how many of them have good or bad data.

Mark Lehmkuhle - Epitel 31:38

Yeah, so one of the advantages of this data going to the cloud at the moment is that we know the health of the session at any point in time on the back end. And that's exactly what Jay's working on right now is kind of a dashboard, if you will, for our clinical care coordinator team who can directly reach out to a patient. And if a session is starting to go off the rails, it coax them back onto the rails.

you know, poor signal quality due to, you know, a sticker starting to peel off or they didn't quite get it right and it got a little too low on the neck and they're picking up a lot of neck muscle artifacts, you know, things like that. They can right now manually intervene, but as you can imagine, we could create systems that automate these processes as well. DeAndre + Grant (32:26.941) I think that's kinda leads me to another question. Honestly, for you, Grant, around, you know, they're solving a human factors problem right now. Right. They we've got the technology figuring out, okay, how do we actually make make it to our user? We use this. Well, I feel like you and the team like face this quite often across a variety of different waking nightmare all the time. Right. Like what what what is what is your kind of anecdotal example? How should you approach that problem? Like if you're kind of starting from the beginning, kind of working.

J Carroll 32:27

Yeah. DeAndre + Grant (32:54.567) Well, my favorite is you need to get burned once or you'll never take it seriously enough. I remember one of our very first times that we just absolutely got cooked, is we had to go past a human factor study for an IFU for instructions for use. And you know, our brains like, How hard could it be? Like you do the the product works this way. And we wrote a set of instructions and we signed off and we absol we performed our own study. Like we we recruited the people, like we did the study.

And you're sitting there and you can't say anything behind the camera because if you instruct them, you've broken the study and you're just cringing. Like that's not how you're supposed to use it. Why didn't you you didn't even open the book? You didn't open the instructions and you're sitting there yelling at yourself. And so I think the, yeah, my first answer, like, how do you actually do human factors design? Well is one, you gotta do wrong once and get get the bruises. And that then is the source bot that will teach you how to do it right. And the only way to do it right is to share your idea. There's no engineering, there's no book that tells you how to empathize with people and how to tell how people do stuff.

You can, you can get closer than like the average engineer who is an introvert. But the only way to really do it right is to test early and often. Take your prototypes, get small user groups. I don't care if they're the same. I don't care if it's corrupt data set of like the same people you test every new product you build with. Just get it out outside of your four walls, the people that physically built it and to your spouse, to your friends, your family, can you use this? And I'm not gonna tell you how to use it. Just play with this. I don't care if it's a medical device or a new keyboard for your c you know, your your phone to type on faster.

Just watch someone other than the people that designed it use it for five minutes and you will learn a million data points at once. Jay, is is this r resonating? I love that. I yeah. Yeah, exactly. Yeah, I was gonna add to that. And I I think if I had a t shirt I could have, it would say the way you do anything is the way you do everything. Like maybe, you know, may maybe that's that's too narrow of me, but I don't know, I really do believe it and and I think this really resonates with me because you know, when I think about how Mark has presented, you know, the the journey of Epitel and and how things got started, we got there because he, you know, we challenged

The status quo, he challenged the status quo. he he thought he could make a better product. And I think that if if you can be humble enough to admit that the first product, the second product, the third product, the fourth product, the fifth product, the end products that you are gonna build, you can continue to iterate on those and you yourself will not get those correct, then you end up basically iterating against yourself instead of just iterating against other products in the marketplace. And I think, you know, when I for the bits and pieces that I've been a part of for Epitel over the last two years, I I believe, you know, the way we do anything is the way we do everything. And that's how we do usability here is that the first time that we build it, we're trusting ourselves that we have the experience and we're drawing on our collective and shared experiences and knowledge to build the product.

But we're gonna build it wrong. But like exactly like what you say, we're gonna get it out there, we're gonna collect some empirical data on it early and often. And then we're gonna challenge ourselves to say maybe we didn't build it right this first time. Maybe there's another better way to build it. Let's rebuild it again. And that product iteration cycle where it's already baked into the culture and baked into the conversation and baked into the process, I I find is just absolutely critical. any company.

And I I'm I'm hope I'm hoping people are listening to this, but any company that thinks they're gonna get it right on the first time, I probably should go find other engineers who have been burned because that's just not how it works. DeAndre + Grant (36:28.777) Well, and to build on that, I think, Mark, you crushed the human factors from the beginning, because you actually set up to solve a human factors problem, not a technical problem. The data you were collecting existed. It just sucked to get. Right. So, you know, your very first you know, we encoded the data into this, you know, field and picked it up at a distance. That just meant there were no wires for the grad students to have to glue back to the mouse's head. And I think this is such a beautiful narrative of like it is just the snake eating itself. It's all full circle.

that you started out to make a human factors problem better, needed to use some technology to do that, which ironically, making the human factors better made the technology better. And it the the circle continues until you build cooler and cooler stuff.

Mark Lehmkuhle - Epitel 37:08

Absolutely. And you know, we're running into this now where it's not necessarily the medical device product itself, but how sites manage multiple studies going on at the same time. You know, this is something that Jay just finished. That's what we call the Remy Portal, you know, that is there to make, to streamline the usability of the system for our customers. Not necessarily the patients who are using it, but for the customers who are

prescribing this to their patients. It's, yeah.

J Carroll 37:39

Yeah, it's a problem of scale or sc economies of scale. DeAndre + Grant (37:39.579) that's Yeah. Well, Deanne, I actually have a question for you, because this hits on something you and I do all the time. Every time you solve a problem, you actually just find the next one. Right? Like e every chain you pull and that snaps, the weekly link breaks. You fix that one and you put it back together and you do it again. And a different link now is the weak link and breaks and it's the friction point. Deanna, you you've when you joined Glassboard, one of the things you did with me and learning how to, you know, mm go to market and some of the like the growth that we were working on together.

Mark Lehmkuhle - Epitel 37:51

Mm-hmm. DeAndre + Grant (38:11.283) Like, Grant, I don't want you to tell me what you think is wrong. I just want to see like the smoke coming out of the gear train when you drive the car. Like you just sat down next to me and like watched how we did stuff and like, that sucks. Like that thing sucks. How do you approach this problem? Cause you're not in what call like technical RD or research. You're all in how do people interact with each other? How do you get a message out? How do you get people to enjoy hearing about a topic, right? And wanting to have continue a conversation. How did you s learn to do this?

That's a really good question. hearing no quite often, I think is probably stuff number one. the the more no's you get, the better you get getting a yes, or at least finding more to get to a yes. no, I think a lot of it, at least on like kind of the more soft skill side, I think when it comes to a product, for example, when you're trying to figure out okay, how how can we make the human factors better? How do we make this experience better for the user that w that way they'll actually adopt it?

J Carroll 38:43

Mm. DeAndre + Grant (39:01.991) I think it all comes back to like behavioral science. Like Roy Sutherland is a guy that I really respect with with Ogilvie over there in the UK. And just that idea of not everything is and Jay, no offense, not everything is backed in data, right? And so from that perspective, like we can look at the data, but people also don't people don't know how to tell you what they want. They'll show you what they want. And so, like one of the best things that you know, from a research perspective when it comes to product development.

Correct. Yes. DeAndre + Grant (39:31.071) is I know the gentleman who kind of figured out the the delta touch faucet, right? And so the the strategy that they implemented was they just created a fake kitchen and or they actually went to someone else's kitchen. This is Thanksgiving and they're watching all, you know, the women and the moms in the kitchen making the turkeys. And they saw the mom walk in with the turkey and they're trying to flip on the sink with their elbow because their hands are full with this big turkey. And they were like, what if we could just what if they could you do the same thing but they could just tap it and they could turn on the sink.

And then that unlocked not just, you know, moms carrying turkeys on Thanksgiving. That's not the entire addressable market. It was significantly larger. As it turns out, no one actually liked to, you know, touch something, the germs or what have you. and so I think that's usually how I approach it in some way, shape, or form of you actually need to go and do actions and see how people respond in the world and at the end of the day, create an experience for the customer, the user, or what have you that really delights them and they enjoy.

I think that at the end the day is like that's my lowest denominator. Not even like, did we hit X goal, but did we hit X goal and everybody that kind of came from it enjoyed it as well? Right. No, I yeah. I think it's the the the experiential part. Right. You you had to watch all the friction that I was experiencing in my job of trying to, you know, talk about glassboard and what we do and every day. And like, you can't possibly do enough of this to be successful. Like you've reached your limit. How do we get a microphone on you, but in a way that makes it feel

normal or good, right? For all everyone listening, like this podcast came out of DeAndre realizing that not enough people knew what we did. So I get to talk with really cool people and half the time talk about what we do and half the time talk about amazing things that other people are building. And everyone enjoys the experience. Right. I mean for Jay and Mark, like it's funny you mentioned that you have, yeah, we're kind of making it better for the patients, but ironically it's the clinical trial administrators we're working on now. They're the current week link. Do you have any idea what the next one is?

Absolutely. DeAndre + Grant (41:24.819) Like in your what I call stakeholder set, not just user set, but stakeholder set down the line from patient all the way up to like manufacturing.

Mark Lehmkuhle - Epitel 41:32

I can think of some. It's not short. So one example is, right now we sell this to private neurology practices and the patient needs to come into the practice to get hooked up and so forth. And so you can see that that is a rate limiting step. You can only see so many patients in a week, yet you probably have a patient population that's much, much larger than what you could possibly see.

J Carroll 41:35

Yeah. The the list is not short. DeAndre + Grant (41:41.107) That

Mark Lehmkuhle - Epitel 42:01

So how do we get this out into the hands of people and allow them to hook themselves up and do their own recording? We also know that a lot of people just can't get to a clinic. Maybe they work two jobs or they're in school and so forth. And so how do we make it easy for them to get these kinds of services without having to leave the house? DeAndre + Grant (42:23.667) That's awesome. Bringing medicine to them.

J Carroll 42:27

Yeah, I I want to answer it a little bit differently from from the software side of the equation and this will probably be like an eye roll comment. but I I honestly feel like the something that is is rapidly evolving, not growing, but just an ever moving target is cybersecurity. Right. And we're always working on cybersecurity. most of my day often is centered around either architecture to support cybersecurity or, you know, chasing whatever is the latest, you know, cybersecurity, you know, threat out there, whether it's, you know, fable or mythos or, you know, whatever frontier like model out there is, you know, Skynet is is coming for us.

DeAndre + Grant (43:14.631) Is it's poking holes in all of your prior pristine IT practices. Exactly, exactly. And I and I I think, you know, the the combination of of that, when we look at especially, you know, bringing it back more to like the hardware side of the equation, you know, I I I've done some work on FIPS one forty two and one forty three and and you know, I think the last time I tried to do this I went to go for like level three, which for your listeners, back in the day, like early two thousand tens or whatever, you know, was one of the the more

higher, more unobtainable levels with US government, very difficult to obtain. And back in the day they were trying to think about how do we secure the hardware in the silicon chips and create proper segmentations such that, you know, if you had to put encryption keys on, you can make sure that they they're not being extracted. And now now we have these other models, you know, like Mythos and and its its lesser counterpart, you know, Fable coming into the equation and we're all terrified of it because

you know, they're breaking loose from the lab and they're attacking other companies. And, you know, I five years ago I just had to worry about a really clever teenager that, you know, wanted to wreak havoc on our system. Now I have this seemingly infinite intelligent terminator robot that wants to come after all of our hardware and our software and I feel like i if you know I don't know it's it's it's

Like I said, it's an ever moving target and there's a lot of people that are just chasing the proverbial cybersecurity dragon around and it's it's a lot of work there. So DeAndre + Grant (44:55.539) Well, I'm so I'm so glad you brought that up because it actually's always been the case. Right. Even before Fable or Mythos were here, anyone developing an IoT connected product, a product with a r radio that could talk to the internet or an Ethernet port, but whatever way it talked to the internet, was actually a cybersecurity risk from the day that it was ever put in the world until the day that it goes in the landfill. And most companies don't plan on lifecycle, right? And life cycle doesn't mean until it dies, it means during the time it's supported.

Yes. DeAndre + Grant (45:26.281) How do you update it? Do you force your users to update and like you have to put usability hurdles in front of them to do so? Right. It's it feels bad when you can't use your device until you go to a software update, but you should brick the device so it can't p stream music from the speaker or it can't act until you update it because you're encouraging bad behavior that isn't about, I don't care if the smart speaker doesn't work anymore because someone hacked mythos on it and it just plays, you know, Rick Rick Astley, it's never gonna give you up all day. That's a fine outcome.

The bad outcome is that Wi-Fi speaker now becomes the hub of a bad actor to get into a house or a hospital or another network. And most people in physical product don't remember it's never done. If you have an IoT connected product until the day your company goes under, that product is never done. You can officially end support and then you can sunset it, tell your users to turn it off. But until you do that, that product needs love, attention, security forever and always.

Exactly. And I think you hit the nail right on the head and a very another important dimension for your users to kind of consider, which is you know, if you're just building, I used to work for Visio, so I actually used to work on soundbars. So this is kind of relevant to me. but you know, if you're building a sound bar and that gets hacked, that's one thing. if that soundbar is the foothold to a larger attack and that is attached to some consumer grade but very critical.

medical device that's attached to your child. And now all of a sudden that otherwise wouldn't have been compromised, but now is compromised. It's really about the risk profile, I think, that a lot of our hardware devices kind of inherently carry with them. Like you said on day one, from the day we've been building product, it's always been there. It's just now they they become more interconnected. That viral spread of that contagion is is now more insidious than it probably has ever been.

And you really have to look at that risk profile to decide, you know, what is the correct posture. Because if you're constantly in defense mode, then you're not in offense mode. And in my my in my brain, offense mode is pushing into the market and challenging some of those trends and evolving the product. But if all you're doing is just guarding your product, you know, you're you're not pushing forward. And so it's it's hard to find the right balance. yeah.

DeAndre + Grant (47:47.848) Well, there's that balance of ship or die. Like you can't make a perfect product. Like as we talked about earlier, like had Mark waited to ship the perfect product, he would have never shipped. Mark, you shipped the first thing that kind of sorta worked, and then you we snowballed from there to get to where we are today with version one, two, three, and so on. I mean, what what version of the iPhone are we on? The first iPhone didn't have an app store. Listeners have heard me say that many times, but like that's my favorite example. Everyone forgets the first iPhone had no app store. It wasn't that great. And and you know, today like half the world runs on it. but this, you know.

Heaven. DeAndre + Grant (48:18.153) How do you balance how good is good enough? How do you balance how safe is safe enough, especially in a medical device, let alone anything else IoT? These are all the decisions that go into making a product. And it usually the great answers to all these choices are a compromise amongst a great team, someone who is hypersecurity focused, someone who's pushing the technical edge, someone who's just trying to turn a profit and you know make cap let the capitalist and machine turnover so we can afford to do it again tomorrow. And it's all these viewpoints that turn over a great product and a great company.

in iteration leads to the future. I mean, Mark, you've started in one side of what I'll call the research and now are firmly in the business side of running a business that does this, that has to survive to continue to do research, to make the next best thing, to make more money, to do it again. How is this resonating from your history? Because it's it's been a, I'm sure, a full circle to get here.

Mark Lehmkuhle - Epitel 49:07

Absolutely. And it's a delicate balance, like you mentioned. So my example would be, when we started this company, there were four founding engineers. None of us had commercialized a product before in the medical device space. So red flags. we're trying to develop this into a medical device that we are wanting to attract venture capital for so that we could then create the commercialization team and take this to market.

DeAndre + Grant (49:21.043) That was a blast, wasn't it? And, you know, everything that we developed so far has been under a predicate 510 K with the FDA. And I got to go through the med tech innovator and learned, you know, depending on what kind of class medical device you are, how you could attract potential venture capital in the med tech space. So for example, if we were an implantable class three medical device, we could attract venture capital without actually touching a single patient. However, on the other.

end of that spectrum as a class two, five, 10 K, it was very clear that to de-risk the development of this product enough for venture capital in this space to be interested, we needed to have that FDA clearance. so Jay's still mopping up the mess on this. So, you know, it basically, we took very little human factors input and cobbled together what four engineers thought the system should look like and do.

DeAndre + Grant (50:33.031) In a room, lo locked in a room together and then it came out. And we pushed it through the FDA on our own and got this FDA clearance. Now, it was not scalable. It was not a product that we could sell. And it had, every time you ran into an error mode, it had a big button that said, session. And people, again, wouldn't read what the modal said and would just hit end session, things like that. And so, you know, we needed to get that FDA clearance to move to the next level. And that was the necessary evil of

DeAndre + Grant (51:01.886) Mm-hmm. know, forgoing proper product development, proper human factors input. You know, just taking the word on four engineers who don't know what they're doing, it was not the best, but it got us to the next level to where we could do that. DeAndre + Grant (51:20.819) And I'm I'm I'm giggling over here because this is I have an analogy about crossing a pond. If you're an enterprise and you want to build a new product, you have enough fuel in the tank for the frog to just jump all the way across the pond. You can just allocate the capital from this enterprise company and you can develop the product linearly the best way possible the first time. When you're a small company, you're a little frog, you can only jump to the next lily pad. And that next lily pad isn't usually a linear line across the pond. It's a little bit to the right and then over there and then to the left again. And this is Mark what you just described as.

Mm-hmm. DeAndre + Grant (51:49.532) I built a product that I knew wasn't the optimal one, but it's the one we can afford to do at the time. And as long as I, well, I'm gonna say it did no harm, right? You didn't hurt anyone building a product, but the product could have been easier to use or more commercially viable on its own, you know, not just as a research tool to get the venture capital to turn it into the version that would be very, you know, commercially successful. This isn't a failure. This is just what we call the next lily pad, which is such an important place in product development to realize that exists. Early founders that haven't played the game don't realize, no.

Mm-hmm. DeAndre + Grant (52:18.995) Just making progress across the pond is the right thing to do. Don't only try and jump into one. And if you drown along the way, call it, you know, that's what happened. It's not the only path. Yeah, exactly.

J Carroll 52:28

I'm stealing that analogy by the way. That's a beautiful analogy. Yeah. Yes. DeAndre + Grant (52:30.963) Please, please take it and popularize it. you know, that that'll be great. I'll like twenty bucks every time you say it and then we can we can we can blow this up. Yeah.

Mark Lehmkuhle - Epitel 52:39

And don't get me wrong, that's super frustrating because we knew that we weren't going to be able to sell this even though that we got the FDA clearance. It was simply a milestone to get to the next level and to the next level and so forth. And you're correct. It's not a linear path. And it just means that we spend more time doing it, but we're still here doing it.

J Carroll 52:57

Yeah. And D DeAnj and Grant, I had a question for you, just based off your experience from all the people that you've interviewed over you know, the the length of the podcast. I I I have a a a personal hypothesis. I'm I'm sure others have thought something similar. I'm wondering if you guys agree with it, which is that if you knew what you knew now about product development and the journey that it would take, you know, to build something, you know

might be a little bit more hesitant. and I I won't be and let me rephrase that. I feel like a lot of breakthrough innovators, like in the case of Mark and and the other founders of Epitel, we are where we are today. I'm wondering if because maybe they didn't know about how painful it was going to be to build a class two medical device or medical device of any kind, but they just jumped in and it was through that desire of innovation and a little bit of the ignorance of the the agency's process

That kind of got us to where we are today. by contrast, for example, I've been approached through my career many times, Hey, do you want to build this medical device? And I'm like, I've done this before. Do you have money? Have you thought of this? And I run through all the reasons why not to do it. And by the time I'm I'm halfway through my list, they're like, Yeah, and and they're they're they're just too depressed of I've, you know, I've I've squashed their dreams. Yeah.

DeAndre + Grant (54:11.19) You're too jaded.

Mark Lehmkuhle - Epitel 54:16

you DeAndre + Grant (54:17.239) I do this all the time. so I I actually have an answer for this. And we can go back, I think it might have been episode two or three of the original podcast where DeAndre even jumped in and we re-randed it. the Hammerhead episode with Rob Martinez and Lawrence Watris. Literally, Lawrence, founder of Hammerhead, a GPS by computer computers Garmin. His quote, I wouldn't do it again if I didn't start with two million dollars. I would have waited till I raised two million bucks to get started to do engineering. but I have a corollary to this, Jay, and I think it's a really good point in time for

J Carroll 54:39

Yeah. DeAndre + Grant (54:46.867) The humans who haven't done it before, and I'm gonna say have ignorance, and this is a good thing, they're ignorant of what's ahead of them. It is the only way they can get to launch a product without the resource of knowledge that they have by the time they've done their first. Everyone that's done their first now is the benefit of knowledge and can throw away ignorance and just do it a different way. But you have to have one or the other to make it to the end. I think this is this, you know, the only way to to learn is to do, and once you know how to do it, you do it a different way the second time. But

Both are equally good for different reasons. I would agree with that. And then on my side, Jay, by the way, that your question to us is the question that I always ask at the end. So this is gonna, you know, this can be I know we're almost at time. So that was perfect, by the way. I loved it. the kind of the the old Una reverse card, the Spider Man meme. But yeah, on my side, I mean, you know, background I I started a company, software company ran it for a number of years. it got acquired in twenty four.

Okay. Sorry. DeAndre + Grant (55:40.599) And the analogy that I would always give to folks is it's like walking up a mountain, but every three every step you see the next three steps. And you don't know if it's you're gonna get to the top of the crest. You don't know if you're gonna hit a plateau. but that's kind of the journey as the first-time founder. And I would often ask myself if I could go back, would I decide to do it again? And not as ignorant as I was, no, I would I was horrible. I mean, every mistake you could possibly make as a startup founder, but I think even

Even after the fact, now that I've done it once, you know, you you if you've been a former founder, people would tap you on the shoulder all the time, hey, would be interested in jumping on this. And I think I'm so much more strict as to what I'm willing to take on and go do. and I just think it's super cool, like a product like your guys's, where you guys are saving lives, right? Like you guys are really trying to solve some really critical problems. Like that's the sort of problem that is compelling enough to say, you know what, like let's go. Let's let's go solve that one.

and so no that that's that's generally how I approach it is when I was when I was young and stupid, I would do anything, you know, just keep on clamping the mountain, this is fun. Now I'm way more jaded and usually say no to everything unless I think it's really that good. I tricked DeAndre really good. I I got I I really pulled the wool over his eyes to get him to join this team. Yeah. I I also love that both of your analogies are both very physical laborious, right? Like it's it's about either jumping lily pads on extremely long ponds or or climbing the proverbial proverbial syphus stairs to nowhere and you're never gonna finish. And I couldn't agree more. It it does it does feel like it's I'll probably spend the rest of my life doing this and I'm not sure I'll ever get to the top, but just keep trying.

Mark Lehmkuhle - Epitel 57:03

coming back. DeAndre + Grant (57:21.715) That's right. Well, Jay and Mark, thank you guys so much for being on the show. Everybody, this is the Hard Tech Podcast, and we'll see you guys next week. Thanks everyone.

J Carroll 57:30

Thank you. DeAndre + Grant (57:32.736) That was an awesome episode. Guys, that was so good.

Mark Lehmkuhle - Epitel 57:34

No?