00:00:00:

00:00:14: Hello and welcome to another episode of the Strategy & Insider podcast.

00:00:19: I'm very happy to be with two really remarkable entrepreneurs who are turning vision into reality.

00:00:25: out is a digital tool that analyzes our voice alone to reveal the first signs of fatal diseases, burnout or cognitive decline before any traditional test could actually detect it.

00:00:40: And reality is technology already exists and applied by major players such as Microsoft.

00:00:47: So are we ready to discover what our voice says about health?

00:00:53: Without further ado... Let's dive into it.

00:00:55: And a very warm welcome to Lara Chavez and Eduardo Giudice, who are the two co-founders of Virtuoses AI.

00:01:04: Lara and Eduardo are really driving forces behind this company which is a spin off from Swiss Federal Institute of Technology EPFL in Lausanne.

00:01:18: French-Moroccan descent and studying microengineering as well as robotics, data science at UPFL.

00:01:25: Her passion for health stems from her career a former professional athlete while her fascination with artificial intelligence drives her entrepreneurial vision.

00:01:35: in that sense she set out to build an AI tool that enhances human connection collaboration but also health by keeping people firmly of the center rather than replacing them with machines.

00:01:48: And back in twenty-twenty, she co-founded her startup Vithals as AI together with Eduardo who himself is a Swiss Italian entrepreneur with the proven track record of scaling tech ventures.

00:02:01: He's a graduate of Bocconi University Milan HEC Paris and Rotterdam School of Management and does hold expertise in economics As well as international management.

00:02:12: His career actually includes roles at the French energy company Total and major bank BNP Paribas.

00:02:20: He's also an ambassador of Geneva Chamber of Commerce, Industry & Services And also what I learned that dedicated Yomi.

00:02:28: So together Lara and Eduardo represent to meet the ideal blend of research excellence as well as entrepreneurial execution.

00:02:37: And with that said, it's truly privileged to have both of you with me today.

00:02:41: and again a very warm welcome Lara and Eduardo.

00:02:44: Thank You for the invitation!

00:02:46: thank you Thomas is great to see you again.

00:02:49: So Lara your journey spans professional athletics advanced engineering and data science studies, as well as founding an AI startup that has already been recognized by the LA Palace Microsoft.

00:03:04: And also others.

00:03:05: you achieved all of this while still in your twenties.

00:03:09: so looking back what was there?

00:03:12: As a defining moment perhaps in your athletic career or early research whatever where You realize that he really wanted to start his own business eventually?

00:03:22: And did you ever contemplate going in another direction since then?

00:03:27: Maybe I don't know enjoying the perks of a corporate

00:03:30: job.

00:03:31: Yeah, so it can be very attractive indeed to enjoy the perks off their corporate jobs especially when your are in Switzerland where they trade-off between cost and life for an entrepreneur compared having kind well-paid job.

00:03:49: Maybe it's not an easy decision for everyone,

00:03:52: but

00:03:53: to me even before I arrived in Switzerland which was ten years ago...I was reading you know as a kid there are these little magazines about science and i was reading one named Science et Vie where EPFL was mentioned a lot about not only the research I think, initially got interested into startups.

00:04:20: And then eventually as soon as I started university... ...I joined a startup competition with friends and we tried building an app in the business model.. ..and they really loved it!

00:04:35: Also on my side of studies i would do some consulting business and different jobs not typical job you'd be doing just as a teacher, or better more like building my own consultancy and developing apps later some data science project.

00:04:52: And so I think naturally i got attracted to building my on projects and So I did try the corporate experience Just because I wanted to kind of leave it.

00:05:04: you see how it is?

00:05:06: In part worked for biotech synthesizing DNA just digital data on it, in the US.

00:05:13: I was with a really nice team and I really liked it.

00:05:16: but after couple of months i realized that I want to have an ownership for my project And be driven by it.

00:05:25: But also at same time I got exposed to U.S mentality and Silicon Valley.

00:05:32: That's how then started working under startup.

00:05:34: Then another one In both cases wasn't passionate enough about what I was working on, but eventually a bit before COVID.

00:05:44: You had to decide in my thesis subject for right after the master and so i started looking into okay how can we make tangible something that is not that tangible, especially for engineer which is our mental state.

00:05:59: Our mental well-being.

00:06:00: How can we analyze it with AI?

00:06:02: And that's how everything started.

00:06:04: and since then I never thought about... Even when it was hard i never thought of going back to a corporate job.

00:06:13: As you were mentioning kind the starting point.

00:06:14: You both have quite multicultural and also international backgrounds.

00:06:19: Where and how did you first meet the two of?

00:06:22: You went out to the idea, to co-found your tools as AI come about.

00:06:27: I would use a word serendipity which is magic war.

00:06:31: in this sense.

00:06:31: it was really by pure art like that life.

00:06:36: so we met really kind of randomly And Lara told me...I was exploring at that time different technology related too To the EPFL Swiss Ferraristo Technology Which is an amazing place to have an idea about what innovation is.

00:06:52: And when Lara told me about the power of the voice and AI at that time, I would like to stress the fact that ShadGPT was not issued released yet.

00:07:06: What really impressed me was just with some acoustic and some mathematics you could spot so many things.

00:07:17: Technology was not there yet, but I was firmly convinced that with a few more research and a few trials we could really build something out of it.

00:07:33: I saw immediately Lara her entrepreneurial spirit in Seoul.

00:07:38: That's basically the things they evolved naturally And i didn't want to let her do it herself.

00:07:47: And then we decided to join the forces in this amazing adventure.

00:07:53: So I really need also watch out what i'm saying and how im seeing it today because you have the potential to analyze my voice, but that said Eduardo You bring credentials from three of the most prestigious business schools Bocconi in Milan HEC Paris and obviously also the Rotterdam School of Management as well a background in scaling tech ventures beforehand.

00:08:19: Among the various projects you are involved and have been involved with, would you say that all this AI is the most ambitious one?

00:08:27: And also the ones with greatest potential.

00:08:30: but if so what makes you believe it's the venture worth betting on?

00:08:35: because when first about the idea that Lara shared.

00:08:41: You were intrigued of getting behind this and doing this jointly?

00:08:44: Absolutely, it's definitely one which is most ambitious because different factors.

00:08:52: First of all AI Is now so fast Compared to any other technology we see.

00:09:01: I would say a cycle of three months in AI is probably comparable to three years in other technological sectors.

00:09:11: It's ambitious because what we try to solve, the problem it really worldwide issues like improving health of billions people and I think that nothing else can be more ambitious than improving life identified condition health conditions diseases as early possible and saving lives, it's probably one of the most ambitions.

00:09:37: And also I would say purpose-driven goal in life.

00:09:42: In terms of why i believe that this is really a company and technology working so hard Is because of the unicity what we are doing.

00:09:56: It enables us to be really one in my humble opinion.

00:10:04: One of the future leading tech company in the world, we are building models acoustic models related to health condition.

00:10:15: similarly two large companies like OpenAI are doing in the large language models and for me betting on this technology given the potentials is betting on the future and hope of humanity.

00:10:33: I know it sounds maybe very visionary, uh i try to be humble because we are a relatively small team and we are still at our initial part of the journey.

00:10:44: but i really believe that we can create something huge with a lot of impact all over the world

00:10:53: And I like the way you position it, and still we come to at a later point in time that your really kind of also already very well recognized despite being still young firm.

00:11:05: But before jumping there probably touching on the core technology our voice as a biomarker, capturing acoustic features such as tone intonation and also speech rhythm to derive insights about a person's health but also emotional state.

00:11:25: Probably for listeners who may be less familiar with the science.

00:11:30: probably Lara asked you could walk us

00:11:33: through how

00:11:35: voice biomarkers work?

00:11:37: what they can already reliably tell us today, where their current limitations are and why is our voice such a rich source of health information compared to other biological signals.

00:11:50: Yeah, so it often surprises people when I tell them that from a thirty-second audio recording of the voice we can assess up to twenty five health condition.

00:12:03: But actually a field of research was born decades ago So actually first publication on detecting Parkinson's from voice recording.

00:12:14: dates form the eighties but The performance that were achieved, we're a bit plateauing until recently when actually COVID and the funding to find alternative way of detecting early signs of COVID.

00:12:33: There is a lot of funding that went into AI models as it's more obvious than respiratory disorders affect voice It's more obviously that you can hear from your voice recording.

00:12:46: University research team started getting funding to do research in that and it also had consequences on the research, other health conditions through voice analysis.

00:12:55: And so why?

00:12:57: It

00:12:57: works basically from the moment there is one of the organs involved in speech production.

00:13:04: So that is brain at length where we have air that comes into our lungs.

00:13:09: And then when we expel that air going through the vocal folds and all of the larynx, um...that produce sounds?

00:13:16: So if you have the brain, the lungs or the facial or the tongue muscles uh..that are impaired Then it would also impact how your voice sounds Even though changes are barely audible to humans.

00:13:32: here An AI model can easily pick them up.

00:13:36: conditions that are, that impacts one of these organs.

00:13:39: It's neurodegenerative disorders like Parkinson and Alzheimer's.

00:13:42: its mental health like depression anxiety disorders.

00:13:46: it also respiratory disorders but also more surprisingly some cardiomystabilic disorders some that are related to chronic heart failure, type two diabetes also.

00:13:58: And there you have either inflammation changing the rigidity of the vocal cords or other aspects in the physiology that changes and impacts the sounds of the voice.

00:14:11: So we analyze all this acoustic feature.

00:14:14: You mentioned some of them like intonation tones, phytorism.

00:14:18: There are hundreds of characteristics in the voice that we can process multiple times per second, every seconds.

00:14:25: And from this acoustic feature there is a model which has been trained on healthy controls and people with these conditions.

00:14:34: So it learns the pattern that is specific like a signature given health condition, And then where we try to be really strong as making sure such models work also on people who have comorbidities because I just mentioned that type two diabetes impacts the voice.

00:14:53: so what if someone has type-two diabetes?

00:14:56: Then surface from depression.

00:14:57: will we be able to pick up depression or only in some she is healthy.

00:15:01: So that's the kind of things we try to be careful with and it tends to work pretty well.

00:15:06: And can I just ask how much recorded voice do you need for detecting this?

00:15:12: Is there something where we talk hours, weeks or sometimes easily detectable depending obviously on the condition?

00:15:20: but

00:15:22: So we need just thirty seconds of speech, and the most important is that it's as natural as possible.

00:15:30: The ideal case is when you pick a voice sample in the background for discussion so there can be teleconsultation or phone conversation between two people who consent to it And this is where they get their best results.

00:15:46: It can also be done with a self-test where you would just talk to your phone, tell about what you did on the previous day or during your holidays.

00:15:55: The contents of what that person says doesn't matter at all.

00:15:59: it's important.

00:16:00: its natural influence in the mother tongue for example.

00:16:04: ideally

00:16:05: That is super interesting.

00:16:07: and if you detect disease I mean which kind of specificity?

00:16:13: Can we detect diseases?

00:16:15: So it varies, of course by health condition.

00:16:18: There are some that are easier to detect like Parkinson's where we can't detect it actually years before the tremor because the tremors is only... It corresponds only one acoustic characteristics.

00:16:31: but then there other ones vary, and they are more subtle.

00:16:35: but we can pick them up.

00:16:37: And in some studies it has been proven that it can be detected up to seven years before the first tremors through such models as voice analysis .

00:16:47: In terms of specific sensitivity on Parkinson's ,we would be around a ninety-eight person or ninety six percent.

00:16:54: so sensitivity specificity But then for other conditions like in mental health, depression and anxiety.

00:17:03: So for example then depression we would be at seventy seven percent.

00:17:07: eighty three percent sensitivity specificity.

00:17:09: so it's lower And most of the health conditions fall within that range In terms performance as today.

00:17:18: Over time We are trying to make sure not just improve that performance But truly make sure it generalize well across languages Because even if you don't analysed languages, it still affects a bit the acoustic especially between tonal and non-tonal languages.

00:17:35: so there's like between Chinese which is tonal languages.

00:17:38: And over German French Italian etc...which are non-tonal!

00:17:43: There we make sure that works as well regardless of accents or language.

00:17:49: You're

00:17:50: super fascinating and very intriguing.

00:17:52: And I'm impressed by the thirty seconds that you already need, also high level of accuracy that we are achieving.

00:18:09: Let's probably dive into a bit different area because for your AI tool language in content is not as important as underlying information they actually carry.

00:18:21: That said some languages must be easier to analyze than others due to its phonetic structure or also cultural communication styles?

00:18:30: Or is there no greater difference that you're perceiving?

00:18:33: and with this, what are some of the technological but also logistical barriers that exist in making technology available let's say all languages around the world.

00:18:44: What I like about our technology?

00:18:47: it really universal.

00:18:48: we since very beginning focused on the paravirbal, no need to understand what people say which also increases the privacy of our technology.

00:19:02: The fact that being a paravirable it makes it universal.

00:19:06: let's say we can definitely identify some cultural communication styles the performance of our models, that's why we prefer people as Lara mentioned before.

00:19:22: We prefer to speak in their mother tongue and avoid any stress during a conversation.

00:19:30: but at same time maybe since there is also some audience in German speaking countries nuance and there is kind of funny story that we deployed some our algorithm with Swiss German.

00:19:47: And the Swiss German had a specific pattern, I myself unfortunately don't speak German but actually can recognize And we start from the, what they call a high German.

00:20:04: Initially our algorithm was performed less well with that dialects but then we trained and used more recordings of Swiss Germans and everything was absolutely in normal standards.

00:20:19: That's the funny story about it.

00:20:20: But generally its performing Chinese Arabic Japanese Spanish.

00:20:27: so really beautiful to see that this technology doesn't have any barriers in regards to languages and genders as well.

00:20:36: It's all about having models that have been trained, evaluated on diverse data.

00:20:42: of course age impacts the voice gender impact states language we just discussed

00:20:50: And one your most compelling loose cases is early detection Parkinson's disease, so through AI-powered voice analysis which was developed in collaboration with the Canadian Research Institute several before and the University of Ottawa.

00:21:06: So actually Parkinson is notoriously difficult to diagnose early particularly because its initial symptoms are really subtle and often miss during routine primary care visits.

00:21:18: How significant could AI-driven voice analysis be in closing this diagnostic gap, which is really a problem?

00:21:26: And what does the evidence gather so far tell us about the accuracy and also the reliability of this approach.

00:21:34: Yes for Parkinson it's one of these health conditions where there are no like single biomarkers that you can just measure.

00:21:44: So traditionally uh You would see how the person walks.

00:21:48: So ideally, you would be in person with them and ask them to walk some specific exercise and see their motor control.

00:21:55: And that's what the input already when the disease has progressed a bit.

00:22:01: then You also asked questions about how they sleep etc.

00:22:06: but Then lot of this symptoms overlaps other health condition which makes it quite hard to diagnose In terms of voice.

00:22:14: so we pick up the tremors, but we pick up also other impairments such as there is one acoustic parameter which is named performance and in particular it's a second one.

00:22:27: It's affected by how you move back part of your tongue so this tend to be impaired already before person has tremor that just one single characteristic when combine all really small change telling about changes.

00:22:45: It's not compared to a reference of the same person, but it is changed compared with norms.

00:22:51: Think for example when you check your cholesterol like there are norm that is valid all humans and then we can be outside of that norm and along with other characteristics in your, or the markers on your blood.

00:23:06: Or else you can check to make sure you are healthy or maybe less?

00:23:11: And so the same way we do voice, We can check independent acoustic characteristic like second formants and be like okay it's out side of norms.

00:23:20: So its to look at.

00:23:22: That is very explainable.

00:23:24: but then under model that predicts whether person has Parkinson or not with a certain probability, it takes into account all this hundreds of parameters to make that prediction and can be really small deviation.

00:23:39: from the norm in many parameters, but then it would tell us okay this person may have Parkinson's and help us detect it early.

00:23:48: And actually we have hospitals that also biotech reaching out to us because they figured out voice is very likely be most sensitive biomarker for Parkinson especially if we want to detect it or monitor it remotely, where we cannot have a doctor in person assessing how the person works.

00:24:11: In addition to long interview but just with a third second not just if the person has Parkinson's or not, but they are getting better with a medication for example.

00:24:21: And as you mentioned also kind of that the stakeholders in health care system like neurologists, psychologists and others how do they react to it?

00:24:32: Are already working with them interested in learning more are somewhat more hesitant because this might be challenge.

00:24:41: what I'm going forward.

00:24:43: How is that perceived?

00:24:44: if you can share anything?

00:24:45: That would be super helpful to understand.

00:24:48: So there was a significant change about, I'd say one year and a half ago so that's late twenty-twenty four where it went from a technology that maybe wasn't super understood, was nice to have.

00:25:09: A lot of people and healthcare professionals get interested into it most likely as everyone started being aware about the potential for AI.

00:25:20: so we had really good perception from doctors.

00:25:27: personal hypothesis is that in the past, a lot of AI diagnostic was around radiology where it would tend to replace radiologist.

00:25:37: And so then there was conflict and not good perception but to support clinical decision, critical rezoning.

00:25:48: I mean if it's your job to diagnose and then also help with treatment patient than you have an additional tool to get more information to treat better your patients.

00:25:57: so they really love its ends.

00:26:00: They tend to look not only at the score okay does a patient have Parkinson or not?

00:26:04: but actually The doctors that use our solution, they love to dip type into the acoustic markers.

00:26:11: That provide explainability but also sometimes I used it to do differential diagnostic.

00:26:16: so we have a really good perception and in terms of their patients So they tend to be willing.

00:26:24: their DP or that they're neurologists use the tool and interpret information along with other information.

00:26:32: They have to make better decisions, so it's really positive on that side.

00:26:37: Cool!

00:26:37: And thanks for sharing this because only if surrounding stakeholders actually work with it then you can make into reality in the end right?

00:26:49: And probably pivoting a bit away beyond Parkinson's kind of diseases that we touched on to here.

00:26:56: Your company is also tackling another major societal challenge, which is the psychological well-being and mental health.

00:27:03: so luckily enough you integrated your solution into Microsoft Teams And it was actually a significant milestone result in your recognition as Microsoft Switzerland style of year back in twenty twenty three and embedding a health and wellbeing monitoring tool workplace platform may create a powerful opportunity for early detection of stress, anxiety or burnout.

00:27:28: But they also raise important ethical questions around how do you ensure that employees feel supported through probably a primary channel for preventative health interventions rather than being watched or surveyed by the employer?

00:27:44: That's absolutely an important aspect of this technology we have to.

00:27:49: First of all, thanks to Microsoft for the support we received and because we work together.

00:27:55: For bringing our technology to their customers especially at that time.

00:28:01: since was twenty-twenty three people were less used to AI tools And we had two explain a lot what it was about.

00:28:10: how are technology and what where they really implication with regards to deployment in a corporate environment.

00:28:20: So we decided since the very beginning to conceive our product with a principle of privacy.

00:28:28: by design, that means each user of Microsoft Teams which wants as to accept the fact of having our tool analysing the call and always have an opt-out option.

00:28:49: So, say that it can activate or deactivate.

00:28:52: this was a first fourth condition.

00:28:55: second condition we implemented is employers cannot get access to individual results but only aggregated data.

00:29:08: And the third one was that employees have full control on his journey and access to real useful insights, and access any benefits or support that a company provides him.

00:29:30: So this is really three ways you tackle these important ethical questions.

00:29:37: We ensure that our technology provides intervention before first symptoms appear and the user, an employee can enter a care pathway as soon as possible.

00:29:51: So maybe to add on that, the beauty of technology is you don't need a specific test for it.

00:29:59: Usually people would do a blood test or some kind of specific tests if they already have doubts like imaging.

00:30:08: also If there are any doubt about having an issue with your health In case of voice because it can be analyzed in the background and from this same voice sample, all the health condition can be analysed.

00:30:23: I mean one that is detected from voice.

00:30:26: then you can screen really in a preventative manner instead of waiting to do it once the person consults because they realise there are some symptoms and some health issues.

00:30:48: Moving probably beyond clinical and workplace applications could also envisage future individuals?

00:30:56: customers, consumers routinely monitor their own boys' biomarkers through let's say a smartphone app or any sort of smart home application similar to how people track steps so hard right today.

00:31:09: And what would need to happen for that?

00:31:12: To be both clinically meaningful at the same time not anxiety inducing into people.

00:31:18: death has always been my dream to be frank since they're very beginning.

00:31:22: although we were there are many challenges too enable debt.

00:31:26: Myself, I've always been using trackers.

00:31:29: I have done my DNA sequencing and like to experiment all kinds of technology on myself provided that I don't know.

00:31:39: what I would like to quote in order to ensure this change happens is a sentence from Steve Jobs as the fact that a technology has to be magic to ensure that it goes beyond either the practitioner, clinicians or corporate sector.

00:32:02: So I would say our technology is magic!

00:32:06: We are almost there.

00:32:08: but what needs to be changed is probably the mindset of people and i think there's really a change I perceive across the population The younger generations.

00:32:21: they're willing to adapt these new technologies.

00:32:24: And also regarding policymakers and decision makers, I've seen actually at your amazing event we participated last year in Frankfurt where you don't see any more silos like practitioners hospital insurance providers corporates consumer goods producers.

00:32:45: now there is a convergence of the health well-being And I think when you have this convergence across different stakeholders and decision makers, together with the open mindset of a younger generation but also late-generation.

00:33:05: We work a lot on what is called silver economy.

00:33:09: we see more people are embracing new technology so if they have the right policy in place preserving privacy To some extent, policymakers push some regulation.

00:33:22: Of course we don't want to over-regulate it but some regulations which we do as in Europe.

00:33:27: actually

00:33:28: What is important fully in line with what Eduardo just mentioned?

00:33:33: And then what is important Is you have the support if something is detected so its really having everything connected.

00:33:42: So where can... You can have that voice check ideally outside of the care pathway.

00:33:48: So it's really preventative and that this checkup is connected to the intervention, so for example what if we detect someone potentially suffers from depression?

00:33:58: It's important.

00:33:59: you have then kind of a solution or support which already there are easily accessible to individuals having all the ecosystem connected together with police, insurance, doctors... And I

00:34:15: love the fact that you bring up that convergence of industries, or the fusion even with industry sectors around one of the most decisive elements in life which is our health.

00:34:27: Our well-being right?

00:34:28: and i'm a firm believer about healthcare in future is going to be much more entrenched into our twenty-four, seven life and not merely only happening at a doctor's place where we are not spending even an hour and half or two per year globally as a person with the doctor.

00:34:46: This Max and at best some bug fixing takes place but healthcare needs to be happening at home that work on our way for it.

00:34:56: so there will need to discuss what you just mentioned.

00:35:00: How can we find subtle ways to ensure that we entrench into healthy living, healthy mobility and subtly gather data for long life?

00:35:11: And then make life better obviously for individuals.

00:35:15: In order be in such a position it is of utmost importance this something trustworthy And that just element is closely also linked to regulation and regulatory approval.

00:35:32: As such, this one of the most substantial hurdles at the same time for any digital health or AI-based diagnostic solution.

00:35:40: Any medical grade AI product must navigate complex frameworks, supranational regulations like those in the EU At the same country specific rules in Switzerland and elsewhere.

00:35:53: What are the tools that I currently stand on the regulatory pathway?

00:35:57: Can you share anything there.

00:36:00: Yes, for us it's at least an exciting part but i think its very important that they already have regulatory approvals required so anyone can kind of trust to help their solution that are proposed to them without necessarily having to deep dive in all the research background behind a given therapy, diagnostic solutions or else.

00:36:29: And so they're virtuosists initially launched as wellness solution and we could immediately offer our technology using vocabulary that is non-medical.

00:36:42: So, for example framing instead of having a depression model.

00:36:47: For example, framing it as moods.

00:36:49: so that's how we launched.

00:36:51: initially in January twenty-three but over year ago...so early twenty five We went through the process of getting certified As a medical device.

00:37:03: with the C marking It can be used in healthcare settings with actually the true names of what we detect and also beyond mental health.

00:37:18: That allows today doctors to use it, for example during the teleconsultation and have actually inside about probabilities displayed with models whether it's an Alzheimer's or Parkinson's depression diabetes.

00:37:35: In Switzerland we had to register with Swiss medics.

00:37:39: but I'd say that most strict regulation is the C marking required.

00:37:46: And for what we do, it's trickter in Europe compared to the US.

00:37:52: So now that we have passed this kind of regulatory hurdle in Europe, We can also easily be available in the

00:38:01: U.S.,

00:38:02: and then further let us say... compliance.

00:38:06: regulatory processes are required only in order to be approved as a standard diagnostic tool without having the needs.

00:38:15: To do further testing.

00:38:16: today, it's a clinical decision support tool so its kind of early warning system.

00:38:21: and then you would do um The gold standard test for the confirmatory

00:38:25: tests

00:38:26: can't be not just imaging etc.

00:38:29: but if we were to be a diagnostic tool than We wouldn't even need to go further, which could happen later on but not yet today.

00:38:39: Especially given the fact that regulation on software as a medical device especially when it's based on AI models is still not fully ready and fully defined their rules.

00:38:53: And one of the common critiques of Europe is that our continent is falling behind in developing proprietary technologies, especially in AI.

00:39:03: You're actually successfully proving quite the opposite with your company?

00:39:08: How do you assess the startup landscape here compared to other regions like US or China and what do you perceive are specific advantages but also disadvantages for building a health tech AI company in Switzerland maybe also with regard to its position outside the European Union?

00:39:26: Absolutely, that's a great question.

00:39:29: In my opinion Europe has been doing quite a lot in recent years but we also work a lot with France.

00:39:37: I would say there is some vision to compete on an extent with USA giants.

00:39:44: We definitely have to be concrete and pragmatic.

00:39:47: so here in Europe we have excellent research centre But we are less keen to take risks, so I think that's the main challenge of European versus American.

00:40:02: You mentioned also Asia and China?

00:40:05: I don't know enough about it.

00:40:07: I'm keen on exploring more for this market.

00:40:12: At the same time, I would say that we really have especially in Switzerland by many other countries.

00:40:20: We are very strong research center where Lara made studies and research like EPFL also of ETH in Zurich Where you can see...I will invite people to go some fair events.

00:40:36: new technologies are coming up from these research labs.

00:40:41: So I think the strength is that we have really strong R&D, a stronger reputation as well?

00:40:48: I would say Switzerland.

00:40:50: when you go and travel across the world where do you go?

00:40:53: to Middle East or even to USA?

00:40:56: Switzerland has taken it seriously.

00:41:00: so there's definitely an advantage in being in Switzerland Models like we do.

00:41:07: at the same time.

00:41:09: The challenges are more on the fragmentation of the market.

00:41:12: so once you quit the R&D stage and You want to go to market?

00:41:18: The market is fragmented.

00:41:21: different languages, different regulation Lara mentioned it earlier which I would say, create a real bar for startups.

00:41:30: And that's something we should really focus and think in terms of policy makers.

00:41:35: but it is very difficult because policies are word written words... ...and mindset needs to be changed.

00:41:45: But personally i am positive.

00:41:48: a lot of change in the younger generations compared to myself when I finished the status already many years ago.

00:41:56: Now there are so many opportunities for European smart students, create their new company and pursue that journey.

00:42:04: And following on what Eduardo just mentioned i think as it started up its slower to start in Europe because of the fragmented markets, also because regulations are stricter both in terms privacy with GDPR and beyond but also medical device one.

00:42:23: But once you're compliant as strict policies then it's easier to expand, because you can basically almost expend anywhere with just some paperwork.

00:42:36: But then in terms of the processes that we have in place... You don't need to change them as they are already compliant with this trickest regulation.

00:42:46: and on the other side another benefit is being born in Switzerland but more precisely it's a multilingual context.

00:42:57: And so for what we built, which is Unvoiced by Omarker then all the data at hand to validate that technology on all these languages and make sure it's also ready.

00:43:10: So

00:43:11: bottom line you have a great technology.

00:43:13: there are emerging evidence full of opportunities in the month or years to come.

00:43:21: If you would have kind of a magic wand and you could change parts in policy, like procurement mindset openness or society.

00:43:31: Whatever it is what are the two three things that do want to change.

00:43:35: first?

00:43:36: To accelerate kind of meaningful adoption of technology more broadly in society at workplace and doctors places.

00:43:44: That's if we say this was challenging question for these podcasts I will say the convergence and inclusion of different industries into well-being, wellness.

00:44:01: On my side it would be most important factor for accelerating adoption.

00:44:10: I'd say a lot of startups criticize regulation in all processes that are in place But I feel quite happy how the landscape is.

00:44:25: And if you go step-by-step, in the end there's a process to navigate and get into markets.

00:44:32: The only thing that is missing are really defined rules on evaluation of AI models.

00:44:40: So for example on voice You would have to make sure... Also enlarge your model.

00:44:46: actually ...you'd need to make it generalize across languages before having it authorised in a country where there is this language that's spoken.

00:44:57: That isn't something we got really asked about, but then for me its not an issue!

00:45:05: We want to make sure our technology works and I think it would be important if you have defined rules.

00:45:15: we had to do processes that we have to put in place, which just kind of waste our time because it doesn't make sense for a pure software solution.

00:45:27: For me its really about having rules adapted to AI technologies.

00:45:34: So really fascinating.

00:45:36: And if some of our listeners are now intrigued by this technology and might ask themselves, how can I actually kind of check up myself?

00:45:45: Is there any chance of doing this?

00:45:47: Can you share on how to best approach the technology in event get a diagnosis ourselves?

00:45:54: Absolutely as mentioned before Our vision and dream is to make it To the large public.

00:46:00: It's not the case yet but we have one possibility to test it, which is our web app.

00:46:07: You can go to appapp.vertroses.ai and we have the well-being tool that you can be tested there.

00:46:20: We are very happy to receive any feedback from your audience

00:46:25: And with that It's a wrap on truly fascinating conversation.

00:46:30: Huge thanks for both of you sharing your vision, science but also passion for building that technology.

00:46:38: That genuinely serves people.

00:46:40: so big thanks to them!

00:46:41: Thank

00:46:41: you very

00:46:44: much!

00:46:44: I personally wish all the best of course and to our listeners.

00:46:48: thank you for joining us on this episode of The Strategy & Inside Podcasts.

00:46:53: as always we truly welcome your thoughts and feedback.

00:46:57: i do encourage Stay tuned for what's coming next and until then take care, stay safe.