#74 - Embracing Tech to Revolutionise Mental Health: Emilia Molimpakis, CEO and Co Founder at thymia

0:00Unfortunately, GPs have so little time, they only have 10 minutes with each person, that they do not have the time to screen for mental illness by default. So you can imagine, like, if you had a tool that worked at the GP level and get it right 90 % of the time, that's a massive improvement in and of itself.

0:21In today's society, there remains a significant lack of diagnosis and support for individuals struggling with mental health issues. Emilia Molenpark is the founder of Themia, a company revolutionizing mental health assessments using cutting-edge gamification technology and AI. Hailing from the stunning Greek island of Crete, Emilia has studied neuroscience to PhD level and coupled with a passion for video gaming and following a profound personal experience when a friend of hers tragically tried to take her own life, brought on by undiagnosed mental health issues. Emilia was determined to make a difference.

1:06Emilia is an online platform that utilizes specially designed video games based on neuropsychology, combined with AI-driven analysis of video and speech data. Her team have developed this groundbreaking clinician support tool which aims to provide clinicians with vital insights kind of bridging the gap between mental and physical health diagnosis. We talk a lot about this. Themia's groundbreaking approach has earned them recognition. They were recently ranked in the top 10 European mental health companies. We cover her journey as an entrepreneur. We cover the differences between raising capital in the UK and the US, the shortcomings of current mental health diagnosis methods within the UK NHS, which I was shocked by.

1:52We also get Amelia's advice to those struggling with mental health issues and much, much more. This is one of the most important episodes that we have ever done. It's certainly one of the most important conversations I've ever had. I think you will take a lot of value from it. This is Amelia Mollin-Parkis.

2:17Amelia, lovely to get you on the Tech Leaders podcast. How are you today? I'm great. Thank you so much for having me. It's a pleasure to be here. Absolutely. Been looking forward to this one. This is a topic that we've been looking to get somebody on to talk about mental health for quite some time. Your profile was very interesting to us. I understand you've been on Radio 5 Live and you've talked about this subject at length and you've obviously got a story of your own. I just wanted to start off with the question that we always ask guests to kick things off, which is what are the key ingredients of good leadership?

2:54And what does good leadership mean to you, Amelia? It's a very good question, obviously, as it's also the subject of the podcast. For me, one of the first ingredients is empathy, which maybe isn't that common, but it's something that has defined me as a person. And so when I became a CEO, it just naturally became my style of leadership is empathy, understanding the people around me so that I can get them to do the best job they possibly can. When they're relaxed and happy, they can do their best work. I think having a very kind of clear set of rules and guidelines internally as to what is and isn't crossable, ethically speaking, is very important, particularly in tech and in healthcare.

3:36So I think those things like a clear vision and what guidelines and rules you want to take as an organization together with empathy that brings out the best and it also creates naturally the type of environment you would want your company to have I think it just naturally occurs as a result sure is does leadership come natural to you I think I've always kind of naturally been a leader in terms of I've always led projects at school I was annoyingly probably the one who was always more bossy and kind of saying what we should be doing what all those things that kind of always had an inclination but it's certainly been a very steep learning curve going from academia into directly being a ceo it wasn't the pathway i was anticipating to be honest and it's definitely been a very interesting kind of learning journey so i had to find my own style i tried very much to adapt to try and be what everybody says in these books that you need to be to be the perfect boss or ceo none of it ever worked for me probably was very much a learning experience yeah absolutely so So let's rewind back then, Amelia.

4:40I know you're from the absolutely stunning island of Crete. I went there last year, as we were talking about in the preamble. I just wanted to ask you about your early days. I think I normally, I just wanted to phrase the question like this. What did you want to be when you grew up? Okay. And can you just talk us through the sort of early part of your life that set you on the path that you ultimately went on? I wanted to be a mixture of things. So after I got over wanting to be a dinosaur, which was what I wanted when I was two years old. I wanted to be a dinosaur as well. Which is a very nice ambition.

5:13After that, I think, yeah, I very much was very interested in genetics, epidemiology, neuroscience. Those were always really, really kind of strong recurring themes. My family ended up talking me out of going into neuroscience. I ended up going into Greek philology instead, which basically means I studied language linguistics and greek history and culture at university so very different but i was always still very very keen on neuroscience and so when the opportunity came up to basically have a specialization in linguistics which was called neuro linguistics or psycholinguistics where you analyzed through language the functioning of someone's brain i found that so incredibly interesting that is fascinating yeah sure i had the opportunity to go to a hospital in greece where i worked with alzheimer's patients parkinson's patients and basically looking at how they use language and seeing how that was different from people same age and gender who didn't have that condition okay sorry can i can i just can i just basically i cannot stop myself because i i heard something before which i was absolutely fascinated by and maybe you can shed some light on it so i heard about the concept of color and I was reading that there was an indigenous tribe of people I think it was in the Amazon who didn't have a word for the in their language for the color blue so therefore blue didn't exist in their in their in their minds in their in their you know in their world so to speak and therefore they just saw it as they just saw it as another shade of green so do you know what I mean so the question is how much does language shape our perception of reality do you think?

6:54Yeah, I think that's a kind of a well-defined language theory, which is basically the limits of my language are the limits of my world. That's what kind of was thought. I think at this point, some people do still believe that. I think mostly it's kind of been found to be, well, you do have a representation of that concept in your mind. It's just experienced slightly differently, potentially. So I think there was also like a tribe who had only variants of black and red in order to express colors but obviously they didn't perceive the whole world as just black and red yeah but it was just different words same with like eskimos that they have 50 different words for snow but you know obviously they'll have a little bit more specialty but it's not entirely determined by that i think what's more interesting is not so much that language determines your thinking but that thinking determines your language so more the opposite way yes of course yeah that makes sense so basically one of the big things i found out really early on is if you want to understand how someone's brain works in general if you want to understand ordered or normal thinking you need to first understand disordered thinking because you can see patterns a lot more strongly there and one of the patterns that i've always used to understand that thinking is language so you'll see a disorder through language predominantly and that helps you understand the brain more generally right okay i just find that I could talk to you about that all day, but we better move on.

8:18We better move on just in case. So, yeah, sorry, I stopped you in your tracks there, Amelia. You were talking about that. Obviously, you got into philology and then that set you on a path then. Yeah, so I did my thesis in undergrad on how Alzheimer's patients use syntax and semantics differently to people who do not have that condition, but who match them in age and gender. And then I ended up applying for a master's at UCL. I got funding to go. Then that kind of set me on the trajectory of neuroscience. So I ended up going into neuroscience and then kind of working, did a master's, PhD, postdocs, looking at different kinds of populations and doing everything from EEG experiments to fMRI to, you know, just brain scans.

9:07but always through the kind of the lens of language use that was always my specialty and that's part of what I bring to Femi as well as that kind of angle of looking at thinking through language essentially yeah sure sure so so talk us through how you ended up in London then did you always see yourself as an academic talk us through that part of your life where you decided to continue your studies and not just go and get a job in industry? Yeah, I think from probably being quite young, I knew I wanted to be a researcher or an academic. I just didn't know exactly what I wanted to research. But I loved learning.

9:45And I loved kind of exploring and finding out about stuff that was always like super, super relevant and interesting. So it was never a question, do I want to do a PhD? It was just what's it going to be on? I ended up going to UCL because I read a paper. Somebody wrote a really great paper on art timers, like syntax use in art timers. I messaged the actual author, something I probably wouldn't do these days, having been an academic. But I messaged the author and said, hey, I love your paper. I love this area. Where should I study? And they said, well, one of my ex-PhD students has just started a program at UCL.

10:24well, she's the best I know in this area. You should talk to her. So I ended up messaging her and she said, oh, we've got a new program. You should apply. And I did. I got accepted and then ended up kind of continuing on with her and doing my PhD with her and a postdoc there as well. I kind of dabbled a little bit outside of that department, going into the Wellcome Trust for a while. But it was never a question, do I want to be an academic? It was definitely, I'll be in academia always. It was more leaving academia that was the new thing, I guess, that was not expected, but ended up happening.

10:59Prior to obviously starting your venture, I noticed you had some experience working on video games. And obviously there's a big neuro-linguistic element to computer games. I've seen pictures, scans of the brain when people are playing video games and things. A lot's going on up there, isn't it? Yes. First of all, how did that video game thing come about? And when was the eureka moment where you saw the opportunity for gamification to achieve other outcomes aside from just playing games? You know what I mean? The pleasure of playing games, you know? Yeah. So I worked with a gaming company briefly and basically the gaming company wanted to have a competitive edge, let's say, over others in the same area.

11:43And they wanted the competitive edge to be driven by science. So you would use science to basically make the game levels progressively harder because it was a word-based game. They needed someone who had expertise in language and psycholinguistics in particular. So they came to UCL. It was kind of towards the end of my PhD, beginning of postdoc. I worked part-time with this company to help them make the video game basically progressively harder. And that pushed me into analyzing other games. And I realized, actually, there are quite a lot of games that use psycholinguistic principles to make the levels progressively harder.

12:21I hadn't seen anyone using it to measure cognition, just to use linguistics basically to drive forward the gaming element, rather than using the gaming element to drive forward an analysis of someone's brain. So I'd seen it one way. I hadn't seen it the other. And at the time, I was still kind of interested in disordered thinking. So looking at things like semantic dementia or Alzheimer's, where you lose concepts, so word concepts. And I suggested to the company, hey, did you realize that if you changed this a little bit, you could target Alzheimer's? You could use it to help people. And they're like, well, that sounds interesting, but we're not really keen to do that.

13:02And I was like, okay. Then my best friend, who was also an academic, kind of went through this whole deterioration phase. She always had anxiety and depression, but I saw her then really kind of go downhill and fall through the cracks in the system. What ended up happening was I realized a lot of the research and all the things we've been finding out about, about the brain and about depression, about Alzheimer's, they're still stuck in research and clinicians aren't being exposed to it. they can't use it. And if only my friend's clinician had these kinds of tools, it could have helped to figure out what was wrong sooner.

13:42You believed in it that much then? You immediately said, right, this can help a diagnosis. It was that clear? It was 100 % clear. I'd thought about it for Alzheimer's. I hadn't thought about it necessarily for depression. But I was like, okay, well, obviously, her clinician must have asked her these questions, or they must have gone into to test her language, test her memory, to test stuff. And it was like, no. He basically asked her on a scale of one to four in the past two weeks, how sad have you felt or have you felt like harming yourself? And obviously she's a very intelligent person. She wasn't going to admit to things.

14:21So he basically said, okay, she's well enough to go home with a prescription for antidepressants. And then two days later, she tried to take her own life. and when that happened we were meant to be meeting on the day and she didn't show up so I I went to her house and and I found her thankfully which is very fortunate also very very unfortunate as an event but it really pushed me and I started to talk to the guy and I said well what happened and he told me you know he asked the questions and she answered and she didn't seem to be in distress I was like surely you use more than that to you know you don't just use questions like no that's what we use i spoke to him i spoke to dozens of other psychiatrists and clinicians and it was just always the same thing i was like but but tech and research has advanced so much more we know so much more we can tell someone has depression from their language use from their face from their movements from their reactions why aren't we using this yeah wow i was like okay well then if this gaming company doesn't want to do it i'll do it that was just it so i just quit and went and did it this just makes so much sense to me because if you if you're a clinician and you've got a patient and you believe that they're just giving you a facade you know they're giving you a story and you can see there's more there i mean asking them to play a game as long as they accept to engage in that activity i mean there's not many other ways you can get under the bonnet and find out what's going on i think it's genius amelia you know i'm sure this there must be others doing this this must be gamification for this purpose must be a thing now how advanced is this and how widespread are people using these techniques so i think it's both common in some areas not so much in the ones we're looking at so there are other companies using games to help treat some mental conditions so embedding cbt for instance within games or using gamification to treat ADHD.

16:23That's something that's been done quite a bit, and it works really well. There haven't been many people doing exactly what we do, which is basically using gamification to feed very complex AI models. So what we specialize in is the games are a medium, let's say, to gather all of the data we need, which spans different modalities. So we gather through our games, which are very simple. We gather voice data, which is from my specialty and also our chief science officer specialty. So there we're analyzing both how somebody sounds and what they're saying. Secondly, we gather video data. So if they're happy for us to switch on the camera while they play, we'll basically look at their facial expressions.

17:04So overlay a grid on their face and look at all the points and see how those move. And eye gaze patterns, upper body movements as well. So say, for instance, twitching, very relevant in ADHD. Slack, facial expressions, very relevant in depression. But also we analyze how you're behaving inside the game. So if we show you a card, do you react quickly to it? Do you react slowly? Do you make mistakes? How do you tap? Where are you tapping? All of those kind of like thousands of data points are going into the models. And that's really what drives the product is the models. The games are a way of getting the data.

17:41so it needs a lot of ai expertise a lot of neuroscience and psychology and linguistics expertise to build these kinds of tools so hence i think it hasn't been used as much even now we know a lot of companies using say voice analysis for depression yeah we've not seen anyone using all three of those at the moment so i think it's it's starting to be used but it definitely could be used a lot more right are the nhs using this these techniques on a significant scale or is this still still very early so i think unfortunately the nhs isn't that welcoming to new techniques and new technologies they make it very difficult to pilot stuff so there is not as much incentive for a small company like ours to expend an enormous amount of resources to get into the nhs obviously we want this to be made available to everyone but it's a difficult process of selling into the nhs you have to sell into every i think they're called icbs now they used to be CCGs.

18:40There used to be 200 of them you need to sell into each one separately. They make it very difficult. I have some hope now that the government's announced, I think it's 21 million funding for AI innovation to be introduced to the NHS. I'm really hopeful, but the private sector is using this a lot more than the NHS is. I've seen other public health systems use this or like embrace AI way faster and much easier than the UK one does, unfortunately. yeah sure talk us about simia yes how did that come about where did that come from talk us through where that came from and the initial part of getting that from concept to product so i think i knew that we could use these techniques like so from an academic perspective it just made complete sense once i saw the gamification element that also kind of fit in really nicely i was like well this makes so much sense also i've been an avid gamer since i was like two years old or so helps a lot I was always like really really keen on games I find them really calming yeah sure but what happened was essentially um all this happened with my friend and within a period of maybe a few weeks I kind of had this urge to do something and to help people like my friend because I felt so guilty about not realizing sooner how bad her condition was and about not acting and I felt like this sense of responsibility towards her.

20:07I couldn't help her at that point. So she was hospitalized for a while and I just couldn't get out of my head that I could have helped her or I could have helped other people. And so it was just a period of weeks during which I was like, okay, how do I stop this from happening again to other people or to her? I need to do this. How do I do it? And so I knew I needed someone to help me build it, but I didn't know how to find that person. and it just so happened that somebody from entrepreneur first here in uh london they reached out to me i mean one thing they do quite well ef is they know where to find people so um they reached out to me and said hey there's a cohort uh coming together do you want to be part of it yeah i was like okay i've never heard of you guys but let me see what you do i was like actually this is fantastic exactly what i need so i joined straight away i was like yep cool I'll apply I got in then COVID hit and we were the first pandemic cohort so just as COVID hit moved to London back I was in Oxford at the time so I moved back to London and I had this really daunting challenge of meeting 100 people on the cohort over the period of two three weeks or remotely and trying to find a co-founder so were the people you were interviewing you were talking to were they all sort of CTOs like coders essentially then is that what you were going for was it more commercial people or a bit of both i when i think when i joined i was looking more for kind of more of the commercial side i thought oh i'll end up basically being the chief science officer or the chief technology officer i don't think it ever crossed my mind i would be the ceo yeah of the company i always thought i'd be a co-founder and etc and then i met Stefano, who ended up being my co-founder.

21:55And we both teamed up in different teams at the time. But there was just something about how we spoke about stuff, how he understood what I wanted to do. And he had all these ideas. I was like, well, actually, if we work together, we can bring in so much more into this than if I was just working with a commercial person. And it just organically evolved into what Themia is today. And so I ended up becoming the CEO and having to learn very quickly what that entailed. EF is a very much venture capital driven company like Y Combinator. Yeah. And it's very much you found a company and then you just get pushed out to fundraise.

22:34Yeah, absolutely. That all happens very quickly, I'm sure as well. It happens very, very quickly. Within three months, kind of pushed out to fundraise, which was way too soon. You obviously raised capital pretty quickly then. entrepreneur first obviously helped you a lot in that journey so you you you had a seed raise is that right have you so where are you at now in your sort of runway you know investment journey whereabouts are you so we've done two pre-seed rounds it's not announced yet but we're doing seed right now essentially so i can't say too much about that just yet but we are doing crowdfunding actually soon as well yeah sure uh top stuff up yeah so we're at seed stage okay so looking back now from when the concept arose right you know so where you are now because you've been on a bit of a journey over the last two years or two or three years what would you have done different amelia i think one of the biggest lessons well two lessons we learned one was don't go out to race until you're fully ready and what fully ready means is very different across the different sectors.

23:37But what we found out was that European and UK VCs don't understand AI particularly well. And so going out to raise as a deep tech company, particularly nowadays, they are looking for commercial metrics. They are looking for stuff to be working, to be proven out. They're very risk averse. Like if you look at the stack, say of risks, they look at you as a pancake stack of risks that pancake stack is massive if you're a uk european vc versus in the us much much lower much more risk friendly i think we just didn't know when we went to raise about that kind of the difference between us and uk eu and we went out to raise sooner than we should have done i think we should have spent more time could you just unpack that in a very high level way was that i wasn't i wasn't aware that i knew there was a difference in terms of risk profile but I didn't realize it was that profound.

24:31I think it's more about kind of the profile, like rather the risk adversity of the funds themselves. So UK and European funds are way more risk averse, particularly in the past year, they've become even more so, obviously understandably with the market, but they see AI as a massive risk profile in itself. So deep tech, much less willing to take a gamble on it. And even if they do, the valuations are so much lower than in the US. And they just look at you and they're basically saying, okay, AI represents a commercial risk. It represents a regulation risk. It represents in health tech, health risk as well.

25:11So basically they're just stacking those risks. And instead in the US, they look at you and they say, oh, AI, you guys are amazing. The potential for this is phenomenal. We're going to back you now before you have stuff going kind of thing, which is very different in the UK and Europe. they don't do that as easily. I think had we known, either we would have gone to the US straight away to raise, or we would have waited to build out a bit more of the product before going to fundraise in the UK and the EU. We learned that the hard way. The second thing we learned, which was, I think every founder finds it out, is don't chase the easy money for commercialization.

25:52If you see an opportunity and it kind of takes you away from what your vision is, yeah don't take it stick with the vision if you have to pivot in order to get like commercial success make sure you're doing that still within the vision you're still you're not sacrificing kind of your end goal for that because ultimately you will pay for that so i think that's something we learned as well yeah so let's go back to the product a bit more then um amelia i'm really keen to understand that the product development journey that you guys have been on can you tell us about maybe when you first started how long it took you to to get to the point of having a ready-to-go product yeah so when we started we basically needed a way to gather data to train our ai models we always knew kind of the models would be the core value driver of themia and games were the ways was was the way basically we would gather the data but we needed a way to host the games so So we ended up building out a full research platform, let's say the Themia platform that has all of the games and all of the, you know, all the tech you need to gather voice video and behavior data.

27:00And we built like, it's probably like the among the best in the world for that purpose, if not the best. And that's what we use to feed the model. So that took a while. Once we had that, the question then became, okay, how do we commercialize this? And this is, I think, where maybe we could have done things potentially differently. If we were US-based, I think we would have focused just on developing the tech, not selling. Because of the way that European and UK VCs and also the market works here, we wanted to commercialize as soon as possible. So we started offering the platform as it was to research groups so they could use it to gather data themselves.

27:40That was the first way we commercialized, which worked well. and then we wanted though to get this into the hands of clinicians so they could use it they could feed us data we make the models better and we help them in the process so we ended up having as the core the themia platform and we built out kind of a full end-to-end sass clinician suite around it so we did everything from booking admin staff to hosting telehealth to note taking and the games fit there so you could assign them to your patients and um this kind of comes back to us seeing there was like clinicians were struggling to do telehealth teletherapy and stuff with their patients so if they don't have a platform for us to plug into how on earth are they going to use the games so we'll build a platform for you that was the first version of the product, essentially.

28:34We did that with a few clinics, but then the more we grew, the more we realized a lot of clinics, particularly abroad, had their own telehealth platform. So they were asking for bits and pieces. So we ended up just giving them modules more than anything. And that kind of just ended up being, well, the more we're breaking down the platform, the more it makes sense to just maybe convert this to kind of just APIs and widgets instead of a full suite. And so we did that in, I think about six, nine months ago, we started doing that. And that's what we're now selling. I think the other big thing we did was we started clinical, like really strongly clinical.

29:13And we only worked with psychiatrists, neuropsychologists. One of the really interesting things haemia does is we analyze all these types of data and we can tell whether someone may or may not have a condition. So we specialize in depression, anxiety, recently ADHD as well. But what we do, which no one else does, is because of the modalities, we can look at the different symptoms that someone has. So I can see whether, say, you, Gareth, may have anxiety. And if you do, how exactly are you experiencing it? Do you have sleep issues? Do you have mood issues, attention problems, etc? And each of these requires a slightly different treatment by the clinician.

29:53So what we've started doing now is we realized symptom tracking is relevant not just clinically it's relevant in a lot of other places so if you have a tool that can track your fatigue levels that's not just relevant when you're depressed it's not just relevant to psychiatrists it's relevant to anyone really it's relevant to any big employer who wants to help their employees it's relevant in safety critical environments so think of places like pilots air traffic control operators minors surgeons it's relevant there so we ended up starting to sell our APIs into the wellness sector as well so that's kind of the next iteration of the product is APIs clinical but then also wellness now as well what about wearables then which capture real-time data on health things I mean sure would that plug into your product definitely yeah wearables are something we've been considering for a while now We already have multimodal data.

30:50So there's a bit of a difference with wearables in the sense that early on, we took a strong kind of ethical stance that we don't want to passively track people because particularly people with depression, anxiety, they feel it's very invasive for them. So we gather data through our games and through actively eliciting speech. Wearables are more on the passive tracking side and it does very much fit in. In fact, our chief science officer has published papers on how do you combine wearables, voice, etc. to track cognition. So it's a natural next step. I think probably we're maybe a year away from that at least.

31:29Yeah, yeah, sure. Can we just dig in very quickly into the user side? okay so i'm just so if you're you know if you and i were having a session and you're a clinician and i'm a patient you would essentially we would have a you know conversation and you would like sort of as part of the diagnosis you would get me to play or to engage with one of these games yeah and then so do you know me as the user would i know that like how how conscious or how obvious would it be that this game is tracking my mental well-being so basically the user journey is if we're talking about the clinical side, say you come to me as a clinician.

32:10Before our first session, I would typically assign you a set of Themia games. Essentially, there's about four or five of them that you might get. And each one takes about two to three minutes to go through. So the whole thing is about 10 to 15 minutes. And you basically just play a few of these games. You know that they're being sent for assessment, but you don't really know what each one is measuring exactly. because we don't want you to know, but you know very clearly that you're being assessed. We make it very clear as well that we say, basically, we're going to play a game now. We need you to switch on your microphone and your camera for this.

32:46Are you okay with that? Do you give us permission to do that? So you're aware. And then as part of the game on the side, you can see your face and you get a sign that basically when you're talking that we're recording. But then the games themselves, they're very basic, but you can't really know what you're meant to be doing so one for instance is a card game so we show you cards one after the other it's a bit like snap so when you see a card that matches immediately previous one you're supposed to tap and when you tap we're recording your face you're tapping here you're not talking so we're not recording speech but that's one example we have another one where you have like an animated image that moves so it's like a cartoon that moves as you talk and we basically get you to describe what you're seeing.