Diabetes Technology Starts: Aurelian Briner from SNAQ on AI-Powered Food Logging for Diabetes

Welcome to Diabetes Technology Report, co-hosted by Endocrinologist David Klonoff from UCSF and David Care from Sutter Health. Hello, this is Dr. David Klonoff. I'm here with Diabetes Technology Reports Starts. This is our new podcast for startup companies. I'm here with Dr. David Kerr and he's going to introduce our first guest. This podcast is for startup companies that have either presented data at the diabetes technology meeting or who have published in Journal of Diabetes Science and Technology. Our guest today from Switzerland has done both. Now I'll turn this over to Dr. David Kerr. Thanks, David. And a huge welcome to everyone to diabetes technology Starts. I'm David Kerr. I'm speaking to you from the sunshine and warmth of Santa Barbara, California, where is our first guest, Arillian Breiner from SNAG, is based in Zurich, Switzerland. Arillian, welcome to the podcast. Thank you very much for having me. It's a pleasure to meet you. Great. We're going to be hearing about your company and what you've been up to and where you're going. But I'm always curious. How did you end up being interested in diabetes technology? Yeah. As many founders in the diabetes industry, my story in diabetes actually also started from personal relationship. In our case, my wife got diagnosed with type 1 diabetes over 10 years ago. I was previously working in software. And yeah, that was really the moment how I got into top of a golf diabetes and hopefully can make a difference there. Great. And what is a fundamental problem that you're trying to solve? Yeah. So with SNAG, we want to make meal time decisions easier or simpler for people living with diabetes. And yeah, the fundamental problem is really around better managing meals. And that's what we are working on for actually close to eight years or about eight years right now with SNAG. And it's still, I mean, even though as you know, very well in the last eight years, a lot of things have been changing in the industry to the positive. And also a lot of things have been changing on the technology side. Yeah, it's a very different place, starting to 2017 when we get started or we are at these days. But somehow the problem of managing meals still exists in one or the other way. And we're still happy to contribute to some of that. Orellen, could you describe what does your product do and how does it make measurements? How does it know what it's doing? Sure. So with SNAG, basically the free components to the product. It is a mobile app where people can lock their meals, predominantly they can lock their meals by taking a picture of their meals and they will identify with different food components. And depending on the camera, it will also estimate the portion sizes. We've done a number of studies, peer reviews that is on the accuracy as well as on the clean color efficiency. I think I see on that. Then the second part, I mean, they're always these days and now you can lock meals, taking pictures, one, you can scan barcodes, you can actually voice lock things. We have a very fine attrition database, and so on. So the meal locking is one part. The second part is we combine that with diabetes data. So you essentially can connect your CGM, your insulin pound, your wearable device to really have, you know, the meal data and context essentially, because all these wearables matter. And then the first piece is more the insights and the coaching aspects that people can get really deep on how different meals impacted their glucose levels, how other wearables around meals impacted their glucose levels. And yeah, helped them ultimately achieve their goals in managing their meals. If someone logs that they had a certain meal, do they have to log what the ingredients are or do you figure out what are the components, how much carbs protein and fat? Yeah, I mean, actually, that's where the latest technology advances came in very helpful. I think compared to when we started, it evolved a lot from being able to more and more automate from, I mean, depending on how accurate you want it, that there can be more or less user intervention, but these days, you get to a pretty decent estimate in some cases by just taking a picture and the output already is like fat protein carbs calories. If you want to have it very accurate, you might want to give it a bit more input or you have some dependencies on special camera sensors, it's phenomenally the amount of automation which is possible. Let's say I eat a turkey sandwich and I take a picture of the turkey sandwich. Do I write down turkey sandwich or does it know that I ate a turkey sandwich from the picture? I mean, if sandwiches can be a bit tricky, right? Because even if I give it to a just a dietician, like you actually, it's just purely from a picture, you would also not perfectly be able to tell what's everything in there. But I mean, it might say that it is a turkey sandwich, it might say sandwich and like propose a couple of ingredients which you don't need to confirm, yeah, it depends. I mean, who is this for? I mean, people would die very enormously, there's no one person with, you know, it's the same as another. So who is this for? And also, is the expectation that people will have to use this forever or do they use it now and again? Or motivation for its use? Yeah. So, I mean, the story originates from the type one space that when we launch into your class, we saw much broader usage, particularly we got significant uptake in the type two market as well. And, I mean, a smaller portion in prediabetes and cessation of diabetes. But overall, I mean, we have a user base of 250,000 people and like 90% of them have some sort of diabetes. So we're really heavy on the diabetes side. Actually, what I would currently at this make more as a denominator is that they use some like glucose monitor, whether it's a BGM or a CGM, that's actually where we are really strong at the combination of meal and glucose and to combine the insights. To some extent, irrespective of the actual diagnosis, we see very similar engagement across the diagnosis times. And to your second question on the usage and frequency, there are really different patterns. What I actually seen also in some of the published studies is that, I mean, the improvements in time and range come very quick. So even after a couple of weeks, you already see an improvement in time and range. And then though, when people stop using, unfortunately, the effect doesn't maintain. I mean, unfortunately, from a perspective of the user, from an industry perspective, yeah, might be more okay. You really need to continue to engage with the product in order to see a sustained benefits. Well, how much of the benefit comes from the picture and how much comes from the logging? It's an interesting question, which we didn't specifically do a study on that. And so, therefore, unfortunately, I can't answer it to you, but for sure, it's something to be like explored at some point. But you provide both and then you combine them. Exactly. I mean, so far, what do we see in the distribution on what type of logging people use, whether they use search or barcode or voice or image? Like image is the most used. So even in those studies, which we published, the image taking part was the most used modality of locking means. How large is your reference library so that you can match pictures with reference foods? Yeah. I mean, this day is it, everyone got a bit got a bit more open, but it's in the hundreds of thousands of on the image recording side. And then in the food database, it's in the millions. But interestingly, also from our observation, I mean, with food logging, I mean, not many items make up like the 80% pocket, basically, of like, they're always or frequently logged. But the 20 remaining percent, there's a really, really long tail, which is in our hundreds of thousands or in the millions of different, like smaller food items. But in the end, you know, you really need to support it the whole bandwidth in order to have a great experience. And here in the United States, food choice is a red hot political discussion at the moment. Major changes in a recommendation. And I'm just curious, looking at all these people with diabetes who have logged, are people are they adhering to guidelines or are they kind of doing their own thing or are people putting in common foods or people deciding here, I don't know what about this weird thing. I've just chosen this restaurant, so I'm going to put it into snack. What's the kind of behavior? What's your impression of people's behaviors? Well, actually, that could be an interesting study question, and we probably would have the data to answer it, but I mean, just, I mean, looking at it, just purely on a frequency scale, like looking at the most like five-time logged food items, it's very common things, like bread, banana, coffee, rice, things like that, which make it up high on the list. So to many surprises on the really on the top picks, okay, I always wonder where people hide if they're eating a lot of red meat or drinking a lot of beer or something, or where they're very honest with snack. Honestly, I think in general, maybe people are a bit more honest with softer, or also if you think a bit further on AI and AI agents and companions and things like that, I think they might be a bit more honest compared to real human interaction, because it's still perceived as being digital. Do you need a person need, or do you need FDA approval for your product to be used in the United States? As long as we don't make insulin dose recommendations, which we don't do, and we clearly disclaim that several places in the app that's not intended to be used for such, we as actually most other diabetes apps as well, which are not bonus calculators, and fall into the FDA enforcement discretion of like lower risk medical and softer, which is to be honest, is a huge advantage for innovators in general. I really appreciate that framework by the FDA, because it doesn't exist in such a way in Europe. Do you feel that your product has some advantages over other apps that people could use for a diet? Well, just looking at the usage numbers, we really have a very significant user base in the diabetes community, and I mean, those individuals figured probably looked at different options in a marketplace and figured that snack might be the one which works best for time. Like where we feel we have a strong edge is really at the combination of food, strong food locking capabilities, and strong diabetes focus, slash the device integrations and insights, and things like that. Compared to maybe in my fitness par, which is much broader or compared to just a lockbook, which is not that strong on the food component side. You're focused on diabetes. Correct. I mean, you're the face of snack. I mean, every time I go to a meeting, you pop up. But just tell us a little bit of the team. I mean, is this a one-man band or do you have the zillions of dollars in your war cherst and you've got a huge team? Just give us some insight into how snack functions on a day-to-day basis. Yeah. We're a really small and nimble team. We're like including external resources, we're six people. Actually, these days, we have with AI and coding agents and all this, we really feel we are more like a 10 to 15 person team because every aspect got so much more efficient and particularly because we are so small team, we can also adapt very quickly. So yeah, we're actually pretty happy. I mean, on a monthly basis, we roughly deal with 20, 25,000 users. So I mean, the product has to be pretty self-explanatory and things need to work decently well that you're able to deal with that amount of users on the monthly basis, that's small of a team. Yeah. And the final question for me is looking back at where you've come from and where you are now, there's likely to be some budding entrepreneurs listening to this podcast. What would be your kind of lesson of the week you'd like to give to people so that they don't make that mistake or they take that big opportunity? Have you got a killer piece of advice? Well, first of all, starting with a problem, you really, really believe is worth solving. And don't fall too much in love eventually with the solution but really more of the problem. Because the journey always takes much longer than what you think and there there is will be a significant roller coaster ride with ups and downs. And you need to be passionate about what you're doing to stick it through, but ultimately or not, it's also rewarding to see the impact you're having that people reach out and say what a significant difference it made in their life and how much they proved on a medical side. Brian, I have one last question for you. Is your company announced any plans for the future? Actually, not too much recently. So what we are right now, the working on, which we released briefly before Christmas, is our what we call coach. It's basically an AI agent specifically developed for diabetes care. So it means that it is able to personalize, it is able to memorize and contextualize basically a very deadly specific and also follow guidelines and regulatory requirements along the way. And it's extremely exciting. We believe there is a huge opportunity for having more care immediately in your pocket. In our case, more like on the nutrition side, like having kind of like a nutritionist in your pocket. For the everyday questions, I don't believe it, I will like, you know, substitute the real professionals, but there is so high demand and sometimes the access is really not there and we really believe there is no opportunity for digital to help those individuals, which might not as easily or not even not able to afford additional care these days. Thank you for speaking with us today. This was an interesting, unusual product and it's a digital health product. So on behalf of Dr. David Kerr and myself, thank you for being interviewed on diabetes technology starts. This diabetes technology report is our first starts. You're the first one and we will be talking to people from other startup companies. A diabetes technology report starts is available at the diabetes technology society website at Apple and Spotify and other sites and we look forward to connecting with you and others in the future. So for now, goodbye. Goodbye. Thank you. Thank you. It was a pleasure to speak to you today.