Julian: Hey everyone. Thankyou so much for joining the Behind Company Lines podcast. Today we have JulianCirco, Co-founder of Hyfe is an acoustic AI that runs on any phone and detectsand classifies any cough. Julian, I'm so excited to chat with you, not only becauseyou have a fantastic name, but also because of your experience as a founder,your background and, and also what's really fascinating about a lot ofcompanies like yourself.

Like we mentioned before the show, we,we chatted a little bit about. How, obviously AI is popularized now withchatGPT and kind of all these other companies working and, and truly pushingthese really intensive and impressive products. But also, it's been around fora while and these, these technologies have been kind of active.

I'm interested in seeing how you'veutilized the technology to really kind of, do something that I, I think, a, alot of individuals don't know, which is really kind of optimizing the, theability your phone has to. Understand your environment, be kind of add somekind of hyper awareness, but also kind of interact with this externalenvironment to really kind of offer a really interesting user experience.

But we'll get into all that in a secondhere. Julian, tell us the audience just for some context, what were you doingbefore you started the company?  

Iulian: Hey Julian. Thanks somuch for having me. I'm pretty sure our name is one of the reasons why youinvited me on the show. And I hundred that, yeah. Yeah, so I'm, I'm what youcall a serial entrepreneur and like many entrepreneurs, I've had a non-linearjourney.

Yeah. So I've started, I've, I can say Ihad like three careers so far. I've had I've had a pretty intense and highlyfulfilling career. As a humanitarian worker. So I've spent 10 years or soworking mostly in conflict and post-conflict environments. So during the twothousands, I've spent time in, pretty much every hotspot that was around inthose days from the Balkans to Afghanistan, to East Timor, to Cambodia, toCongo, to Somalia.

Very educational journey. Of course,I've learned a lot. And ironically, I've learned a lot about entrepreneurshipbecause in environments like that kind of low resources, low infrastructureenvironments people tend to use to be very creative in the way they usetechnology, in the way they use what they have at their disposal.

So I've learned a lot about that andabout just, just ways that technology can be leveraged to achieve impact atscale. Yeah. And so far then I've Go ahead, go. Please continue. No, no, and,and I mean, then I've had another career in just, just running kind of largecomplex operations in frontier markets.

So I've spent many years in sub-SaharanAfrica. Running, building products sort of trying to, trying to open marketsfor impact products. And that kind of launched me into building my owncompanies. And I've built a number of them over the years, across different,across different different kind of categories from consumer products toFinTech.

And now MedTech.

Julian: It, it's sofascinating thinking about, how other countries adopt technology and describe,for the audience that doesn't know kind of the difference in how technology isin, in other countries or in hostile environments. Being that, for example,India was, was first able to kind of bring, come online with the mobile ratherthan on web.

And so how they've offered such adifferent experience on how they view consumer products and people engagementand the types of things that they built. What were the things that you saw interms of the adoption of technology and other countries and what were some ofthe, the, the applications that were popular that are just different from whatwe see here in, in at least in the US.

Iulian: What I, what always,what always fascinated me is how, like if you think in the us a phone is a nicething to have and it's super convenient. But it's not life changing in anysignificant way. It's just, it, it changes the quality of our life, but itdoesn't affect our life to that extent.

Whereas in a place where you don't haveaccess to infrastructure, you don't have access to information, it becomes. Akey component of, of the way that you manage your life, whether, in those days,this is before the internet, right? So you, we're talking early days of, ofcell phones, early days of sms.

People used to use SMSs in ways that inthe west we would've never thought about it. Like, everything from, fromgroups, people who cannot distribute groups. People would, would shareinformation by SMS to some sort of person that would then disseminate it usinganalog means like, Down to, to the level of having like a blackboard withchalk, right?

Yeah. So you receive, you receive, thinkabout things like you receive information on the price of, whatever produce in,in the next district. Mm-hmm. And then the guy who receives it, he's in a teahouse or whatever, and he has a blackboard and he writes the prize on theblackboard, and people in the community come and see and say, oh.

The, the mice or whatever is cheaper inthis district than in the other district. And then they make plans accordingly.If they sell maze, they try to get there. If they just, it's, this is the thingthat I've kind of, I, I've been impressed by and just learning about how people.

Transfer some of the, some of the socialvalues into the way they use technology, like social networks and things peoplewould vouch for each other using phones across, state borders and, in, in,across what would be a conflict zone where you could not physically go and, andsee that person.

Yeah. And that in a way that, lookingback, that kind of launched me on this. On this journey where I'm trying tofind ways to leverage technology in ways that haven't been possible or, or, or,or haven't been used or, or, or contemplated before,  

Julian: yeah. Well, it's sofascinating thinking about the speed of the communication and also the type ofinformation that they're, communicating with.

It's not just, a conversation, but it's,it's information. It's, it's, downloading of, whether it's ideas or groups or,like you've said, the vouching system, which is fascinating because. A lot ofus think about, our social network and how that essentially vouchers for us as,as we broaden our expand, who are connected to who our mutual connections are.

But you know, I guess what's differentabout the direct, kind of value of, of, being vouched for someone or havingthat kind of tool to be, so readily available in a conflict zone, for instance,how key is it to, to, get information quickly and use it and then, continue tokind of adopt and adapt?

Iulian: Yeah. Yeah, andthere's another side of it, which is, I mean, I, I think it's a bit moreintuitive. And that has to do with the fact that where you don't have legacy infrastructure,it's easier to leapfrog to the next generation of technology. Like if you, ifyou have, like in the US you have all these, all these legacy, pieces ofinfrastructure, but even pieces of sort of contractual infrastructure, what youhave with Verizon and T-Mobile and things, it's really, really hard to leapfrogover that because.

People are locked into this contract andthey're sort of, there are investments that are made that are kind of hard toget out of at the level of the companies, but also at the level of the users.Whereas if you think of a place like India, there's none of that. So then it'svery, very easy to be agile around.

Using the new wave of technology, evenif it's not that reliable. Even if you don't have them. If you don't have them.The sort of the packages that you'd have in the US and eventually leading theway with innovation.  

Julian: Yeah. And what aresome ways that you've seen in terms of the advancements of.

Of, of at least cell phones or in thistype of technology in phones that are allowing at least companies like you, ofcourse, but also other companies to really utilize the, the technology inthere. Being able to sense things, be able to monitor, having the, obviouslythey have the I always forget the, the name, but the essentially the, thesensor that allows you to understand.

Stand where you are in the world, almostlike a compass. Right. What have you seen in the advance of technology and howphones have been really used to go beyond just communication? What are somethings that you've seen outside of obviously what you're working, Hyfe and, andwhat you're  

working on?

Iulian: I mean, I think thisis, this is a huge opportunity. You, you have basically very high performancesensors. That you have on your person all the time, wherever you are, whoeveryou are these days, right? Yeah. It doesn't matter if you are, if you're inBurundi or in India or in Los Angeles, you have a device on you, at least onethat has a chip set that is more powerful than any computer, 10 years ago.

Yeah. And this chip set is connected toa bunch of sensors, which are super precise. They're consistent, they're lowpower. It's amazing. And actually I'm amazed. I don't see, I don't, I'm amazedwe don't see more innovation in this space. Right. And initially you see that,like with sensors and with data in general, you have these waves that are, thatare, that are kind of the same.

First you have the op, you can collectthis data, but you don't really know what to do with this data. And actually,People who are kind of, they're vested in a particular industry. Take, takehealth data, for example. They're very suspicious of it, right? They say, well,this is just noise.

I don't know what it means. I getJulian's heart rate, and now I get Julian's, exercise rate and how many stepsJulian's made yesterday, and it's just too much and it's very noisy and I don'tknow what it means. And I can't rely on that in any way because I've beentrained to take. Decision on information that is of a different nature.

Right? Yeah. I asked Julian how youfeel, I take your temperature, but with ai, that noise separating the signalfrom the noise becomes so much easier, right? Yeah. Yeah. And this is what,this is like, like a thesis that I have as an entrepreneur and I'm betting onthis, on, on this thesis, is that we will be able to handle a lot of noise andfind signal in a lot of noise.

And I want to build companies and I wantto invest into, into products that, yeah, operate in that space because I thinkthat's kind of the next wave, whatever it is. Whether it's imaging AI or soundai or just, just shifting through a lot of data. This is, this is for me one ofthe most compelling use cases for ai.

You can just consume this huge amountof, apparently, noisy data and find signal in it. It's super exciting.  

Julian: It's, yeah, it's all,and thinking about just like the innovation piece that you mentioned, why, whydon't you think companies are innovating more with neo cellular devices andusing the sensors and sophisticated?

Is it, is it privacy? Is, is it challenging?Is it difficult for users to adopt? What do you think are some of the reasonswhy companies aren't necessarily innovating? In, the, the mobile space, is itless attractive as an entrepreneur if, if you don't know it, and what do, whatare some kind of common reasons that you think company or founders are, aremoving away from, innovating in that space?

Iulian: I think, I thinkthere's no playbook and that can, like, entrepreneurs, like to run with aplaybook. And I, I respect that because there's so much unknown that you mightas well minimize the, the surface area where you're taking risks, right? So, sowhat you see a lot now is playbooks, people just applying playbooks todifferent kind of angles, whereas, The opportunities that I'm talking about aremore like blue ocean, it's more like no one's ever done this before. So thenyou have to, you have to deal with innovation maybe more than a typicalcompany. Right? Like take, take my company hi. Right? Like we are. We're on, weare basically decoding medical signal out of sound. That's basically,fundamentally what we're doing.

Yeah, but what's the playbook? We don'thave one. How do we charge? Can I do a SaaS about this? It's, not really who,how do I package this as a, as a product? How do I, like what's the, like, thevalue is very clear, but how do I deliver this value to a user, and how do Iget them to turn on their microphone?

And there's just so many, there's somany points on the journey that you have to innovate. Which means that the riskis higher, which means that there are less people playing at it. But we're, Imean, what I've noticed over the last three years that we've built in thiscompany is that we are slowly defining the playbook and we are kind of the,we're the sort of company that building the open.

Like, we, we kind of share ourexperiences and stuff and I'm excited to see a lot of companies that now applyour playbook. Yeah. And I mean, I'm familiar with several of them and I'mexcited about it because I think that, I mean, we are, we are. There's enough,there's enough opportunity for, for more of us, and there's always gonna be adifferentiator.

But what's nice is that slowly we'redefining the playbook for this. And then you, you have a playbook that we kindof pioneered in, in health sound, but. It can be applied in other areas aswell, yeah. That's quite,  

Julian: yeah. Well, obviously,just to describe the technology a little bit more, I would, would love for, foron, for the audience's context as well, but using it essentially san to decodemedical data.

Kind of what was the inspiration behindthat idea and how have you been able to use the sensors and, and, and obviouslyI have some more questions about that, but please, define the technology alittle bit more for the audience.

Iulian: I mean, when you thinkabout it, when you think about it, there's. Like imaging is a huge category,right?

Like look like, let's just focus onmedical imaging. Everything from x-ray to MRIs to laboratory work is basicallyimaging and AI is proven over the last 10 years to be very, very powerful inthat finding patterns that are consistent with cancer or with some otherdisease, so yeah, it's a non-controversial.

Big opportunity and there's thousands ofsuper innovative, clever companies playing in there. Yeah, yeah. But, butimaging has a few issues. Imaging, you need specialized lenses, you needspecialized light, you need specialized spectrum, like you need infrastructurethat is, that is medical to work with imaging.

Right? Yeah. But sound is verydifferent. Sound is, Is continuous sound is not, you don't requiresophisticated infrastructure for sound, and our thesis was, well maybe there'smedical, maybe there's medical information, medical signal in sound. And youask doctors, Hey, do you use sound in your revolution?

They use it all the time. They say, canI listen to your lungs? They say, could you please, cough, they say, and whenyou think about it, our bodies produce a lot of sounds. Yeah. And we are, weourselves without any medical training. Not some sound, we can tell bad soundsfrom good sounds.

Right, right, right. But our ears, amongall our senses, our ears are super low resolution. Right? Yeah. Like we don'thave good hearing as humans, but machines are really good at hearing and yeah.And machines are really good at processing sound data. And now all of a suddenwe have. We, we are, we are unlocking all this signal in health sounds and wefocused on, I mean, you could build, I think you can take our playbook andapply it across any number of sounds.

Everything from snoring to sneezing, tobabies, crying to, but we focused on cough because the most, the most amazingthing to me as a non-medical person is that. Cough has the way that we evaluatecough in our, in kind of clinical settings hasn't changed in hundreds of years.Yeah. Like if you go now, Julian in the best hospital in LA and say, look, I'mcoughing.

You are back in the 16th century, butJulian, how bad is your cough? Well, doctor is pretty bad. You, you. That'sbasically the level of evaluation. There's no way to quantify it. There's noway to describe it. You don't have a language to describe it. So we thoughtthis is a great opportunity. So, and we kind of committed to building tools tosupport coughers across the journey.

Yeah. All the way from first symptoms totreatment. And it turns out that there are hundreds of millions of coughers outthere, people who just cough and, and. More importantly, there are coughers whodon't understand why they cough and there's no, yeah, there's no treatment forthem and there's no diagnostic for them.

And these are our people. We just buildtools for them.  

Julian: Yeah. It was sofascinating, like how it's able to interpret, but also, when, when you kind of.Go and, and categorize the different sounds. How do you go about that process?Do you go and, and do you apply them to a medical board? And how do you reallykind of go through the categorizing process to, then understand what coughshould be categorized as in, in a certain, like, a subset of potentialillnesses versus, asthma or any other kind of medical condition.

How'd you go through the categorizingprocess?  

Iulian: So, so we are doingsomething counterintuitive, Julian, and it's one of my favorite stories isthat, yeah. When you start categorizing sounds, listening to an individualsound and trying to analyze it, you are back into imaging limitations.

Yeah. You need to control the quality ofthe microphone. You need to control the acoustic environment. You need to makesure that the sound is crisp and high quality. But that's not how we operate.Because, if, if we do that, you might as well go to a laboratory and do a bloodtest. Yeah. But what we've learned is that.

cough frequency, which is the amount oftimes you cough over a certain period of time, contains huge inf chunks ofinformation and frequency is easier to collect than high quality sounds becauseeven if it's bad quality, you know it's a cough. All you need to do is binary.Is that a cough? Yes. No, and that's where we put our efforts.

We are the only company that buildstools for longitudinal cough tracking. And it turns out that, I mean there,meanwhile, over the last few years, there's been a lot of research and,clinical trials and peer-reviewed papers that legitimize our approach becauseit turns out that there, there are consistent.

There are consistent pattern Yeah. Onyour, the way that you cough over time, and I'm not sure if we use video forthis, but I can share with you some incredible images that you can just look atthem with the, with the, with just the eyes of a layman and see, whoa, I cansee a pattern there. Yeah, go ahead.

So by just. Yeah, when you think about,when you think about cough frequency, think about like an, like an ecg, anelectrocardiogram, right? The rhythm of the heart. It's the same. You havethese patterns that where cutting those up and down and up, and if you cancorrelate those patterns with things like environmental factors, like, we canbasically.

We can basically detect things like athreshold, a humidity threshold for an asthma sufferer, or for a C O P Dsufferer, which is in plain English. Say, oh, if you, if the humidity in yourroom is above a certain percentage, the risk for an asthma kind of episode goesup five times and, or we can, there's some interesting research that I, I I canshare with you that shows correlation between cough frequency and outcomes incovid patients, meaning that, When people, and this is also counterintuitive,when people, people's cough frequency goes down, if they're in a criticalcondition, right, that predicts negative outcome, they're likely to beintubated or to die.

Which is, which is, goes against thecommon sense cuz you think, oh, they cough more. That means that they're,they're sicker. But there comes a point where if you cough less, that's areally, that's basically an alert sign. And you can use these sorts of. You canuse this sort of information in very low infrastructure, cheap, easy to scalesettings that change, potentially changes the way that we think about treatingrespiratory disease.

Yeah. Super exciting stuff.  

Julian: Yeah, and, and youmentioned something in terms of like, obviously defining the playbook, that wasone challenge, but also, another challenge is adaptability and who essentiallyis gaining the most value out of a product like this? Obviously there'sconsumer, I think, value that, that we can all gain by understanding ourenvironment, both, at home and kind of broadly if, if we're out in the openkind of.

Care to identify, certain costs incertain areas in particular. But on the institutional level, it what is,outside of, doctors keeping track of patients and things like that. What aresome creative ways that you've seen in kind of more institutions? I utilizethe, the information that you're kind of collecting and understanding more soto kind of help with patient outcomes.

How have you been able to navigate notonly the adoption from the consumer side, but from institution side and, andwhat were the challenges that you faced there?

Iulian: So there's a lot ofinstitutional opportunities, right? And you can think about like, they'rebasically two big, big opportunities with institutions.

One is public health, and there we playa, we, we play a significant role and particularly during the the Covidpandemic, there was a lot of excitement around what we do because car frequencypredicts. If you, if you think about cough frequency at an individual level,that's one thing. But if you think about it at the aggregate level, like you,you take your cough frequency and then you, you aggregate it with yourneighbors and with your neighbors neighbors, you can actually see heat maps ofcough frequency.

Yeah. Think of it like a ways for cough,right? And you can see where cough frequency goes above the baseline, right? Sothen you're like, oh, in this particular neighborhood there's an outbreak ofsomething, there's some pathogen. And you can think in terms of neighborhoods,but you can think in terms of campuses, you can think in terms of buildingswithin campuses.

You can think in terms of the, the southwing of this particular building. And that allows you to manage to basicallydetect pathogens in real time, which is a bit of a, it's like a very, very hardproblem historically. Yeah, because, how do you know there's a pathogen? Cuzyou have to first know that there is a pathogen and it's like, it's a very hardproblem historically and we cannot solve that.

Just by having real time coughfrequency. We don't know what the pathogen is, but we know that if a lot ofpeople start coughing more than normal, something's going on in that area. Sothat's kind of the public health. And as it happens, a lot of us in the team inthe Hive team have a public health background.

So it's, it's an area where, we feel alot of affinity with, there's a lot of work that we do in TB and in other kindof public health respiratory areas. The other one is, like pharma and justinnovation. Sure. Particularly around, around medical innovation. One thingthat blew my mind is when I found out that there hasn't been an an antitussiveapproved by the FDA since the late fifties.

Yeah. 1958 or something. Now, do youthink we reached peak calf science in the fifties? No, I don't think so. Sowhat happened is, Between the fifties and now, what changed is our scientificstandards. Like now, these days you have to generate evidence that is veryrigorous right? Around, you have a claim, you have a new drug, right?

Right. Your claim is, well this drugwill help you stop coughing. But that ev generating that evidence is reallyhard if you can't measure cough. Yeah, right. You can't say, Hey, you can't gothe regulator to the FDA and say, Hey, look. I've got 10,000 patients who claimthat they cough less because they took my medication.

They will just throw you out. You can'tdo that. Yeah. And because you can't measure cough, you can't generate thatevidence, which means there hasn't been any innovation in in cough medicine in70 years. And that's crazy because if you start coughing now you have twochoices. You either get a placebo drug, some sort of syrup or something.

Mm-hmm. Or you get a highly toxic drug.Or both. You get basically something that gives you a lot of adverse events andit just, placebo, it doesn't actually, doesn't actually work. And by bringing,by bringing an ability to measure calf and to understand frequency, we areactually building ways to customize management of calf, right?

Like we can see how whatever you do tofix your cough, whether it's you controlling external variables, youunderstanding what the triggers are, or you are following some sort oftreatment. We can see in real time if that treatment works or not. And then youhave, for the first time ever, you have input output type dynamics when youmanage cough, which is, it's something that should happen.

It's crazy that it isn't in 2023, butjust coughers don't get the benefit of data-driven decision making in, in, intheir clinical setting,  

Julian: yeah. You know, beingthat is such an, innovative, technology and, and you're, in a lot of, kind ofalmost like a, a blue ocean of, of opportunity.

What are some of the biggest risks thatyou face, as a company then, if, you kind of see the, you're building aplaybook, you have the roadmap, you're getting a lot of adaptability, what'sthe particular risk that you see in term, whether it's external or internalwhile you're kind of building towards this eventual future.

Iulian: Oh, so funnily enough,I see very little risks with our products. Like, I think the technology sound.I think our thesis, our assumptions are proven, I mean, meanwhile we have. Wehave global scale. We, we have tens of thousands of people using our, our technologyevery day. We have big corporate partnerships.

So, so the product and the market, whichusually would be the big risks innovation, are, I feel very confident aboutwhat I think where our risks are, are in, in this fa in, in the fact that weactually don't, we are kind of category defining. So then when you talk toinvestors, you have to convince them that there's a category there, which is.

Very different from saying, look, I runa SaaS and here are my numbers and my month-to-month growth is this. And theycan extrapolate in their ex Excel sheet what your valuation is. If I go andtalk to an investor and say, look, I'm building this blue ocean thing, andthere's a lot of people that are excited about it and here's our traction, buthow do you evaluate that?

How do I get, like the investorparticular at the beginning is gonna give me five minutes of their time, yeah,how do I use those, those five minutes to get them to see the opportunity? And,and some of these, some of these challenges are quite interesting, but I enjoythese challenges because it's, it, it, it, it forces us to kind of simplify ourthinking, to focus on what's important.

So, yeah, I actually, bring it on. Like,I, I, I, I am, I'm happy to have this sort of problem rather than, oh, no onewants my product, or, oh, my technology doesn't work, yeah, yeah. So, so we arehaving small, we're having small, I mean, we're having. Good, good problems tohave,  

Julian: yeah, yeah.

Even just thinking about, the pitchingprocess, I, I, I haven't, you're probably the first founder I've asked, but youkind of, where do you start the story? Where do you start the journey whenyou're communicating the value of the product? Do you communicate with, theevidence that you find, or, the personal journey?

Do you communicate by that ability ofthe technology? Like, where do you start when start, starting to hear aboutthat journey?  

Iulian: Yeah, it's a hardquestion. This meme with the. With the bell curve of intelligence, you knowwhat I mean? Like that. So think about that meme and on the left you havefirst, oh, be very simple in your communication.

And it's the, basically the steam. Andthen you have like the high iq, be very com, be very nuance. And you com. Andthen you go down to the Jedi and it's be very simple, yeah. That's kind of the,that's kind of where you have to be. Is, is like, it's like, And it's a, it's aprocess, right?

So, so what's interesting is over thelast year, something happened in, in the World, which is just AI became soobvious, the value of ai, and it's became so obvious how fast you can buildthings with AI that we haven't, we haven't, we, we know we have a change in ournarrative, which is beneficial to us. Before a lot of our narrative was builtaround the power of ai.

Hey, we build these AI tools that findsignal in the sound. Now, I don't actually have to build an narrative about AIbecause I can finally focus on the value that we're creating. And this is thefirst time in this company that I can do that as, as someone who pitches theinvestors. So I, I see that as a, as a, as a huge.

There's a huge progress piece ofprogress for us because I don't have to sit in front of an investor and tellthem, look, AI can find patterns in sound. They get it, and actually they'reexcited because they're following. Investors are very much following trendsjust like everybody else, right? They're following the trends, so, Now I canstop talking about ai, which is a hard thing to talk about.

Yeah. And just talk about value. Andthen when, when the conversation comes to how are you delivering this value andhow does your technology work, that means we are already passed the point wherethe value is obvious, the opportunity is obvious, and we are in a place wherewe can go into more than nuance conversation.

So I'm, I feel that the big trend andkind of the popularity of AI is working again in our faith.  

Julian: Yeah. Yeah. I guess,if everything goes well, what's the long term vision for  

Iulian: Well, I mean, we areredefining the way that we, that you, you approach cough at the moment, yeah.Which means that we are defining the way that you approach respiratory health,which means that we define way, the way in which you think about health, and weare kind of unlocking this whole, this whole category of signal, which is soundpaste.

It's very exciting.  

Julian: Yeah. I love this nextsection. I call it my founder faq. So I'm gonna hit you with some rapid firequestions and we'll see where we go. Yeah. Cool. Let's do it. I always like toopen it up. What's particularly hard about your job day-to-day?  

Iulian: Well, I mean, I'm, I'mpart of a team that's very scrappy, so there's a lot of things that we're doingand I think the hardest thing is finding the focus to do kind of deep work.

It's really, really hard. Yeah. Yeah.But it's part of the game, yeah.  

Julian: What, what's somethingthat you, you least like doing and what's, something that you, really enjoydoing part of in terms of part of the business?

Iulian: I mean, there's a lotof admin overheads that I really don't like doing.

Yeah. Like, I love five minutes meeting,I love canceled meetings. Yeah. It's just like, it's very, very hard to, toallocate my bandwidth and I think everybody in the early stages has thisproblem. Yeah. And you realize that bandwidth is one of the few zero sum gamesthat you're playing, and. Is super important to prioritize, yeah. To be kind ofruthless in, in, in, in your prioritization.

Julian: Yeah. Yeah. Thinkingabout, one thing you said earlier is, is the icp, the individuals that youknow, really kind of find the value out of that when, when you communicatevalue to say a consumer, to be able to say, adopt this technology.

Allow it to use your microphone. That'ssomething that's at least. In the US that, that, a lot of individuals strugglewith is how much of my information do I really care to opt in? And howfrequently do I, care to be opted into, having an app collect information. Howhave you been able to kind of surpass that and communicate simply the value of,understanding not only your, your personal kind of cough, but also yourexternal environment, how that affects, how have you been able to build thattrust?

Is it education? What vehicles allow youto build trust with your consumers?  

Iulian: No, I mean, I thinkthe best way to build trust is, is by just building trustful technology. Andlike we are, for example, we are, we're running all our models on edge now,which means there's no data living your phone.

And that's good enough for most people.But I will say something that's kind of a hot take on this. Yeah. Which is the,the friction that comes with, oh my god, someone's listening to my phone. Whichnormally would be super dangerous when you build a consumer product. It worksin our favor because it's a segmentation filter.

Yeah. I know for sure that people whowho bother to download our product and use our technology really need what weare providing, yeah. So I, this is how I make sure that our users are allcoughers, right. As opposed to just random people who are curious about whatthis is all about. And. This is something that in the long term works in ourfavor because we work with a, we work with a highly motivated segment ofpeople.

And yes, initially our growth was a bitslower, and yes, initially we had a few issues, but that means that we knowmore about coffers now than anyone. I can make this claim with full confidence.We know more about CERs than anyone on this planet, and that includes largepharma companies because we actually have a lot of coverers that we talk toevery day, and we.

We adjust the way that we, they interactwith our product and our technology based on what they need to see and how theywant to use that in their daily lives. And that's very valuable to us as acompany,  

Julian: yeah, yeah. Thinkingabout, the capabilities and, and building out a playbook, outside of whatyou're doing now to, hi, what, what would you do the playbook for, if youweren't kind of keeping track of coffers and, and doing everything around coughand, and public health, what would you use the playbook for in terms ofutilizing the sensors and mobile if you weren't working on Hfe right now?

Iulian: So I think that the,the biggest part of the playbook that I would u utilize again and again is thefact that we have managed, no, AI is based, like the quality of your AI dependson the quality of the training dataset, right? Yeah. If you have a, if you havea bad dataset, your AI will, and this is a huge challenge and actually a reasonwhy very few companies play in sound AI in particular, because you see.

An l lm or sort of like a generalpurpose, AI can be trained on data from the internet, which is, which is huge.Enormous. Yeah. But sound. How do you train on sound? Like what? How, how doyou train your first model on sound? You have to create this sound, and thenwhat happens is your typical team will go to their friends and say, Hey,Julian, please generate this sound for me, calf once or whatever.

And then your first model will be builton a highly biased. Data set. Yeah. And it'll work in the data set, but themoment you take it in the wild is just gonna be crazy. Like those early, earlygeneration of Twitter bots who just became racist after like three days. Yeah,it's exactly the same problem.

It's there's, there's inherent bias inthe way that you train your model. We've done something completely different,which is we have created value for consumers from day one. Yeah. And againstthe advice of kind of senior experienced people, we built a consumer product,although we had no plan of monetizing that, and in a very short while, we hadpeople in 150 countries.

Yeah. Downloading our product and usingour, which means that our training data set is now 700 million. Samples ofsound. Yeah. Collected over 150 countries collected in every possible acousticenvironment you can imagine collected across every social demographic segmentyou can imagine and collected on every combination of hardware, software userbehavior, which means that our model works in the real world.

With a very, very high performance. Andactually I can send you a link where you can test it without any optimization.It just works on your machine, whatever your machine is, yeah. Straight in thebrowser. Yeah. Yeah. It means that we are ready for scale and we are ready forthe real world,  

Julian: yeah. Yeah. What,what's one what's one thing that you. You haven't seen your per, hype datautilized that you would like to see it outside of obviously public healthindividuals knowing themselves. But what's one kind of use case that hasn'tbeen necessarily prevalent yet that you're particularly excited about?

Anything come to mind?  

Iulian: So, I'm very temptedto tell you because we are actually working on something. But I have to becareful to not tell you too much is that, we, we, there's been a lot ofinnovation. In, if you think about the, if you think about the, the journey ofsomeone who gets sick, there's been a lot of innovation on the sides.

Like, oh, you, here's a better way tocommunicate to your doctor. Here's a better way to, yeah, send your data to thedoctor, but fundamentally, very little. Has been disrupted in that space. Youstill have to travel to see some person who's very busy and they're gonna giveyou 15 minutes.

Yeah. And already that's slowlychanging. AI in particular is gonna change that. Then you go to diagnostic, youstill have to go to a laboratory and take a sample of data, which is. Which istrue on the, on the Tuesday at 9:00 AM when you took it. But I, it tells younothing about what happened before or after.

Right? So bringing longitudinal datathere is what we are doing and it's super exciting, right? Because now thediagnostic, the, the durability to judge what's going on as a doctor with apatient evolves with the way that that patient evolves every day or every houror every minute, right? And finally you come to treatment where, For hundredsof years.

Yeah. The treatment is, here's asubstance, please ingest this substance. Right? Yeah. What software can changethat? Yeah. What if we have a way to use less of this external substance thatyou're taking, which has a, is designed to work in a kind of a broad spectrum,and what if we build something custom?

Because now we know how Julian's bodysounds, which means that we can infer what's going on inside there. What if wecan build some sort of custom treatment for Julian that works only for him, butit works so good that he's gonna be healthier sooner at a lower cost toeveryone? And I think that's possible.

And I won't say more, but. It's anexciting, it's an exciting opportunity there.

Julian: Yeah. It is excitingto see how sophisticated the information can be after a certain, obviously,amount of data is collected and how, how, uniquely tied to an individual it canbe. And it really kind of opens eyes of, of the possibilities on how to.

Not only kind of utilize the tools andtechnology we have, but also the information we have and, and kind of marrythat too. So, so excited to see kind of how that, that's been, you've, you'vemade leaps and bounds of progress, at least in, in sound and AI and coughing,but how much more that can, you can extrapolate that across different use casesand things like that.

But I always like to ask this nextquestion just because I love how founders extract knowledge out of anythingthat they ingest. If it was early in your career or now if you think early inyour career now, but what books or people have influenced you the most?  

Iulian: So I, I mean likeeverybody, I had my face when I was consuming a lot of, of management booksand, I mean, I think that there were a few early on that had a big impact onme.

I mean, the Blue Ocean is one of them,must be like mid two thousands when this book came out. But I think thatthere's, I mean, I think that there's a lot of, like, it's a space that hascreated a lot of noise. In fact, we probably did an AI to extract the usefulinformation from all this content that's been built to advise startups.

But I think that you can boil likeprobably the most impactful thing I've done and I continue to do, and I'mtrying to keep doing it, is just ship. Yeah. Like it's, it's a stupidly simplething, but it's very, very easy, particularly when you work with sophisticatedtechnology with innovation.

It's very easy to get sucked in tryingto over-complicate things or trying to over-design them. So, so I just boil itall down to just, if you can ship every week, you good. And even if you shipsome monstrous. Quality in the first few weeks, you're gonna learn very, veryquickly what to improve, yeah, exactly. So that's kind of my, that's kind of mymantra, just like, like go out there with like, very, very low level of qualityor, or fidelity and just keep improving.

Julian: Yeah. Yeah, yeah. Iknow we're at the end of the show here, so I always like to make sure that wedidn't leave anything on the table.

Is there any question, Julian, that Ididn't ask you that I should have, or anything that we didn't talk about thatyou'd like to touch on before we we end the conversation today?  

Iulian: No, I mean, I'm justsuper happy to, I'm, I, I really enjoyed this conversation, Julian, and yeah, Imean, I'm, I'd be super happy to to provide details, connect with anyone whotakes an interest in what we're doing and, try to see how I can provide valueto the community.

Julian: Absolutely. And, andJulian, I I love to always ask this at the end, where can we find you? Give usyour LinkedIn, your Twitters, your, your websites. Where can we not onlysupport you as a  

Iulian: Yeah, I mean, I,obviously, I have a, I have a LinkedIn you probably will share the link in the,in the show notes.

I'm on Twitter, , @i_circo. And yeah, Imean, I'm, I'm looking forward to connect with anyone.  

Julian: Amazing. Julian, it'sbeen such a pleasure. Yeah. Yeah. It's been such a pleasure. Learning aboutyour early experience and also kind of what, what are the interesting, excitingways that you're innovating and with H, and how you're really able to offer adifferent depth of information that it's really kind of been missing,especially with.

Previous documentation, previous practices,but tying technology into it to really, increase the effectiveness,effectiveness and also the speed at which we can understand the inf informationoutput of not only ourselves but our external environment. It's only gonna beexciting to see you grow more and, and see what kind of other technology you'reable to kind of integrate and adapt on.

But it's been such a pleasure chattingwith you. I hope you enjoyed yourself today. And thank you again for joining uson Behind, Company Lines  

Iulian: you so much for havingme. This was a lot of fun.  

Julian: Of course.  

‍