Building Pet Health Tech: ML, Sensors, and Dog Behavior Data | Sofya Yulpatova

The transcripts are edited for clarity, sometimes with AI. If you notice any incorrect information, let us know.

Sofya's pet tech startup with machine learning sensor data and behavior pattern analytics

Alexey: Hi everyone, welcome to our event. This event is brought to you by DataTalks.Club, a community of people who love data. We have weekly events and today is one of them. If you want to find out more, check the link in the description. You will see all the events in our pipeline. (0.0)

Alexey: Do not forget to subscribe to our YouTube channel so you get notified about future sessions. We also have a great Slack community where you can connect with other data enthusiasts. The link is in the description. Sign up and you will get an invite to Slack. During today's interview you can ask any questions you want. (6.0)

Alexey: There is a pinned link in the live chat for submitting questions. Click the link and we will cover your questions during the interview. Now I am stopping the screen share. I am opening the document with questions. We can start. (41.0)

Sofya: Sounds great to me. You took a sip of your tea so it seems you are ready. (1:01)

Alexey: I already showed this to Sofya but I want to show it again. This is a souvenir I got at Web Summit, a conference I attended recently. Was it two or three weeks ago. It feels recent. The event was huge and very impressive. (1:08)

Sofya: It was almost a month ago. Time flies. (1:19)

Alexey: This conference in Lisbon is enormous. It was my first time attending. I had heard it was big but I could not imagine the actual scale. The main stage is like an arena where large concerts could be held. There are also five or six pavilions with exhibitions. (1:29)

Alexey: I attended only on Tuesday and managed to go through one and a half pavilions during the entire day. That was all I could cover. In one of these pavilions I met Sofya. Hi Sofya. That moment was memorable. (2:01)

Alexey: I was walking through the stands and noticed something related to fitness and measurement. Since I use a Garmin watch and enjoy running, I like tracking health metrics. I saw Sofya at the stand talking about tracking. I said that I also care about tracking my fitness data. We talked for a while and at the end she said, You know this is for dogs. (2:18)

Alexey: I also got this pin from her. For those listening without video, the pin shows a happy dog on a leash. The dog is smiling and wearing boots. I wondered why it has boots. I also asked if it is a he or a she. (2:49)

Sofya: It is a he. His name is Milo. (3:16)

Alexey: So Milo is smiling, wearing boots, and the bottom says Fit Tails which is the startup Sofya is running. The design is fun and memorable. (3:22)

Sofya: That is correct. (3:34)

Alexey: This is a longer intro than usual but the day at Web Summit was intense and full of impressions. All those memories came back while I was talking. Today I am speaking with Sofya whom I met there. We will discuss sensors, technologies for measuring dog behavior, and how to analyze that data. Sofya is the founder and CEO of a PetTech startup. (3:40)

Alexey: We have FinTech and EdTech and now also PetTech. You can describe her startup as an Apple Watch for dogs. Her work sits at the intersection of machine learning, sensor data, and real world behavior patterns. We will discuss the challenges of animal health technology compared to human wearables. I know something about human wearables from using my Garmin watch. (4:17)

Alexey: I have no idea how things work for dogs so I am curious to learn more. This curiosity is one of the reasons Sofya is here today. I was impressed by what I heard from her at Web Summit. Sofya, welcome to our event. Now you can finally speak. (4:52)

Sofya: Thank you, Alexey, for inviting me. I still have many emotions from Web Summit. It was my first time attending. It is the biggest gathering of startup and tech people in one place. It is huge and full of energy. (5:16)

Sofya: You meet so many people and have many conversations that the emotions stay with you. It was wonderful to have the chance to attend. I already bought a ticket for 2026. There was a discount so I decided to get it early. I am excited to visit again. (5:35)

Alexey: Okay Sofya, tell us about your career journey so far. What did you do before, and how did this all lead to creating a startup. (6:13)

Sofya: This is a difficult question because my background is a mix of tech, product, and doing many things at once. Migrants often need to juggle many responsibilities. I will not go too deep into personal details about moving. I recently moved again and that added to the journey. I moved to London in September so I am still new here. (6:23)

Alexey: And you are originally from Latvia, right. (7:01)

Sofya: I lived in Russia, then moved to Latvia, and now to the UK. It has been a long journey. I have been into tech from a young age. My first serious project was a create your own bouquet tool where you drag and drop flowers. I was maybe 13 or 14 when I built it. (7:01)

Sofya: I was amazed by it and felt very proud. I thought my whole career would be in programming. I found the project years later and it was terrible but that is how everyone starts. I had good teachers who encouraged me. I loved going to Olympiads in English, literature, and math. I often skipped regular lessons to attend competitions. (7:41)

Sofya: That is how everything began. I joined the math and informatics program at school. I won the school Olympiad and went to the regional one. Then I went to a boot camp to prepare for more competitions. At 13 or 14 I was mostly doing web programming and making small sites and games. After school I enjoyed working on these projects instead of homework. (8:32)

Sofya: I loved programming because you could build something alone from scratch and focus on details. After school I went to a full time job and chose the coding path. Later I did go to university. I worked as a programmer for a while. Over time I lost the creative spark because the work was focused on delivery and speed. (9:21)

Journey from programming hobby to full time software development career

Sofya: I realized coding in a company was not for me. My hobby had become restrictive. I wanted creativity and depth. That is when I moved toward product work. I wanted to analyze the market, shape strategies, talk to people, understand problems, and build products. That felt more aligned with what I wanted to do. (10:00)

Alexey: I had a similar experience in my first job. At university everything felt exciting, but at work I had to fix bugs in an old enterprise Java project written by someone who left. It was draining and boring. When a hobby becomes a job, it changes everything. (11:13)

Sofya: Exactly. In my first job there was no documentation or processes. There was also the story about the genius developer who left a week before I joined. I had to figure everything out on my own. This was around five or six years ago. It was not a great experience and I knew I wanted to switch to product work. (11:29)

Sofya: I moved into product and project management. At a small consulting company I made the classic mistake of helping with coding. I thought I could quickly fix a few things and ended up doing everything. The team relied on me and management pushed me to do more. I learned the importance of boundaries. (12:27)

Sofya: I left that company and joined an automotive parts company. That job was like a founder boot camp. We built an ERP system for automotive businesses from scratch. Their previous system was a mix of spreadsheets spread across Google Drive and OneDrive. It was chaotic and confusing. (14:37)

Sofya: I talked to many clients and mapped all their processes. I was not into cars at all and did not even drive, yet I had to understand everything. I interviewed logistics, sales, marketing, and every department involved. I built a team and enjoyed working with them. When I joined there were five of us in development. (15:35)

Sofya: It felt friendly and collaborative. I later scaled the team to 18 people. Eventually the environment became less cozy and we needed to focus on strategy and bigger challenges. I still miss the team because they were great people. We solved hard problems in short timeframes while dealing with a ten year old legacy system. Some days I worked sixteen hours and it was intense. (16:40)

Career growth after skipping university and building practical experience

Alexey: All of that you did without any formal university education, right? You finished school and went straight into work which is amazing. I have not met many people with a path like yours. Usually people study something unrelated and then switch to tech and their journey is similar except they spend five years at university first. During those five years you could have progressed your career already. (17:20)

Alexey: Often people who start after school and people who start after university end up at the same level. In Germany I met people who went to a full stack boot camp right after school. It lasted four months and they soon found a job. They skipped five years of university and replaced it with four months of learning which is an interesting career choice. Later I checked your LinkedIn and saw that you did attend university. (17:26)

Alexey: I remembered myself at seventeen when I was finishing school. I did not know what I wanted to do in life. I liked programming but also liked chemistry, yet there was no chemistry program in my city so I went with programming. Many people at that age feel lost and choose something because parents say it is good. They try different things, get work experience, and then understand what they actually want to study. Was that your journey. (18:04)

Sofya: I would say not exactly because if I had wanted to study product management I would not have chosen computer science. I am not even sure if one can study product management directly at university, maybe business administration is close enough. For me it was simple because I had a hobby I loved and I wanted to learn it deeper. That is why I joined university. I combined work and university at the same time and it was quite a journey. (18:49)

Puppy adoption story and family influence on pet focused innovation

Sofya: Some university courses were frustrating because with real experience you realize things do not work exactly as described. Other courses really helped, even for product management, because they gave insight into how everything functions. For me going to university was mostly about academic studies in machine learning. I initially thought the entry barrier was too high to try it randomly the same way I learned web development. At that time resources were not as widespread and university felt like the simple path to access structured knowledge. (20:07)

Alexey: The reason I wanted to talk about your career is because your path is interesting and uncommon. It is a cool story. Now I also want to learn more about your startup. How did you come up with the idea. Tell us. (20:50)

Sofya: It was a radical decision because I always wanted a dog. I never adopted one because as a kid parents put pressure by saying there was no time and as an adult I understood the real constraints. I moved a lot, worked full time, studied, and everything had to be done in person. There was no room for a dog even though I wanted one. (21:12)

Sofya: At some point my job became more remote so I only had to go to the office two or three times a week. I adored working from home and thought it might be time to consider adopting a dog. There is also a funny story. My husband and I were scrolling a website where you can find various things. In Latvia it is ss dot com which is similar to many listing platforms. (21:52)

Sofya: He was joking about adopting a spider even though I cannot even look at one from five meters away. Then we clicked on the puppies section and saw that a Labrador puppy was available. It was the first of January and we wrote to the breeder asking if any puppies were still available. She said four were left from a litter of ten. We asked if we could come and see them. (23:00)

Sofya: She suggested meeting the next day but my husband insisted on going the same evening. He said if we waited until tomorrow we might not adopt anyone. We ended up driving an hour and a half outside the city in freezing weather. We met all the puppies. One chewed my boot, one ran around, one barked, and one sat in front of me and looked into my eyes. (23:49)

Sofya: We spent an hour outside at minus fifteen degrees and we were freezing. We chose the puppy who kept looking at me. I felt she chose me and still think so. We took her home and everything changed the same way parenting changes your routine. A puppy needs constant attention and it completely shifts your day. (24:41)

Sofya: It is two months of sleepless nights because the puppy is stressed and cries. She needs to go outside often because she cannot hold it. This was a real journey. The sad part is that the breeder was not honest with us. Our dog had a passport and documents but we did not know what tests should have been done. (25:57)

Sofya: She did not show one specific test related to hyperkeratosis. We found out later that it is an inherited allergy that causes the nose and paws to crack and bleed which leads to infections. It was terrible because we were anxious and I felt guilty for not noticing earlier. We visited vets two or three times a month and still did not know the trigger. I track everything about myself, so the lack of data for my dog frustrated me. (26:48)

Sofya: I wear an Oura Ring and an Apple Watch and you wear a Garmin so you probably understand what I mean. I am obsessed with data and at some point wondered why I had nothing similar for my dog. I researched available devices and bought one that looked good. It showed steps and calories which was interesting but did not solve our problem. We bought a second, a third, and a fourth device, and our dog ended up looking like a Christmas tree with trackers around her neck. (27:54)

Sofya: None of them worked for what we needed and that is when we thought it might be worth building something ourselves. That is how the journey started. The existing devices collected basic data but did not offer tools to explore it deeper. They focused on simple and fun metrics to keep users engaged which is fine for subscriptions but not enough for real insights. They did not explore the underlying behavioral signals at the depth needed for early health detection. (28:40)

Sofya: The early signs of health issues appear in how states evolve over time. Sleep fragmentation, subtle restlessness, changes in variability, small shifts in movement quality, and variations in movement quantity matter. This is more like anomaly detection than simple classification. You need a learned baseline for each specific dog before detecting deviations. Dogs differ dramatically from one another even within the same breed. (29:39)

Dog health monitoring framed as anomaly detection in real world machine learning

Alexey: I imagine tiny dogs bark at everything because they are scared and everything around them seems huge. Big dogs act like nothing matters. Even size alone makes patterns different and then breeds have inherent diseases so you need to factor that in. And on top of that every dog is unique. Your Labrador and her sister can have completely different personalities. (32:04)

Sofya: Exactly. One can run ten kilometers and hide from people while mine is lazy and lies on the couch until she hears food. It is not only about personality or eating behavior but also about the mental stimulation needed to feel healthy, physical activity needs, and sleep patterns. It is similar to people because we all differ. I need at least eight hours of sleep while some of my friends feel fine with six. They can have interrupted sleep and stay functional while it ruins my entire week. (33:04)

Alexey: I can relate to that. What kind of data do you collect. You mentioned movement. Is it like steps or something else. (34:34)

Sofya: The device collects IMU data which includes gyroscope, accelerometer, and magnetometer readings across all axes. If we talk about activity labels we collect time spent active such as walking, playing, and running. We also collect sleeping patterns which are complicated. The collar device sits under the dog’s neck and tracks its orientation and movement. It is similar to what your watch does when tracking your steps and other activities. (34:42)

Sofya: The device tracks its position in space which is not three dimensions but twenty seven. By analyzing these patterns we can detect what the dog is doing and what is happening. That covers activity and sleep. And sleep fascinates me a lot because I am a data driven person. (36:21)

Collecting canine data with emphasis on sleep patterns and cycle tracking

Sofya: We track the sleep cycle. For dogs it is important to account for both night and day sleep because both affect energy. We track how much they sleep and how their sleep looks, including whether they move a lot during sleep. (37:05)

Alexey: That is with the same device, right? You do not track heart rate here and only look at movement. If the dog stays in one position and does not move, that may indicate a deep sleep phase. (37:35)

Sofya: We would love to explore heart rate variability. For people optical sensors provide rich heart data, but dogs are furry and some breeds are like giant snowballs, so we cannot ask owners to shave them. Our new device is very sensitive to breathing patterns, which are easier to detect because subtle location changes occur when the dog breathes. (37:52)

Sofya: We are still working on it. Distinguishing heart from noise in IMU signals is difficult. Runners use chest straps or different sensors that detect heartbeat more directly, but chest straps are not practical for many dogs due to fur and comfort. (38:50)

Sofya: The core problem is that IMU sensors can capture small movements that may correlate with heartbeat, but fur and different heartbeat patterns across breeds make reliable extraction hard. For very furry dogs it is complicated to detect the heart signal, whereas for lean breeds like Dobermans it is easier and you can observe distinct patterns in sensor graphs. (40:03)

Alexey: Is heart rate data necessary? (41:21)

Sofya: Not always. While heart rate is essential for human health trackers, vets use different patterns to identify issues in dogs. Sleep is our second-most important metric after weight. Weight is visible but not early enough to catch problems. Sleep can reveal pain or issues two to three days before activity changes appear. (41:27)

Sofya: Owners cannot easily observe nocturnal awakenings or sleep twisting, so sleep metrics provide vets a rich source to assess pain or whether further tests are needed. This is why we frame dog health monitoring as anomaly detection based on a learned baseline. (42:29)

Establishing a dogs normal baseline through long term data observation

Sofya: We cannot build a baseline in two or three days; it takes two or three weeks to account for environmental distractors. People, weather, a new family member, or changed routines affect behavior, so we need time to learn what is normal for each dog. (43:35)

Sofya: Once we learn the dog’s normal behavior, deviations help detect anomalies, but we must account for aging and contextual changes. It is a continuous journey. (44:32)

Alexey: Can you detect something useful from day one using population-level models like Garmin does when it asks age, height, and weight? (44:59)

Sofya: Yes. We take into account breed, age, and weight and use population models for initial classification tasks such as counting steps, estimating calories, and coarse activity labels. But anomaly detection requires learning the individual baseline. (45:41)

Sofya: After a few weeks of data we can detect if something seems off for that specific dog. That is the core requirement for useful health alerts. (46:14)

Alexey: You said a few weeks. Earlier you mentioned about three weeks. That makes sense. I wanted to ask about your transition to full-time startup work because I also left a job to focus on a project. How did you decide to leave and focus on the idea? (46:33)

Sofya: It was very difficult and I know it is hard for everyone. I felt my time at the daily job was wasted because I wanted to allocate it to something important to me and my partner. I lived with that thought for more than a year and was terrified to leave stability. (47:20)

Sofya: I was attached to my team and the social stability; we had many activities and a safe zone. The alternative was to pretend it was not important, which felt worse. Financial reality matters too: you need money to live, and London is expensive. (48:18)

Startup funding through personal savings and early stage bootstrapping

Sofya: I had savings and my husband supported me, so I took the leap and pursued Fit Tails full time. We chose not to take investment before shipping a product because raising funds would take about a year, whereas we could bootstrap and solve the real problem now. (49:34)

Sofya: We did spend money on prototypes and faced hardware troubles; some of that was wasted but we learned a lot. Eventually we will need investment to manufacture devices at scale because producing physical products requires capital. Kickstarter works for some cases but larger production needs funding. (50:36)

Alexey: With software you can bootstrap more easily, but for hardware you need capital. What prototypes did you build and how did you design them? Arduinos are large, so how did you create the first device? (51:49)

Sofya: The first version used samples from AliExpress; we assembled components to see if we could capture useful signals and understand the data. Even clumsy prototypes provided initial insights. My dog was the tester from day one and adapted to wearing the collar because dogs already wear collars for walks. (52:26)

Sofya: Initial prototypes were heavier, but our current device is around 36 grams, which is negligible for dogs above three kilograms. We soldered early units at home and disassembled commercial devices to identify components such as ESP32 modules. (53:26)

Sofya: We realized scaling requires more reliable parts and manufacturing, so we worked with an outsourcing company and faced mismatched expectations, timelines, and budgets. It was a learning journey and not always pleasant, but everyone goes through it. (53:56)

Alexey: Everything you went through to get here is fascinating. I am excited to hear more about what comes next. (55:03)

Finding cofounders and collaborators through meetups and coworking communities

Alexey: You keep saying we. Who do you mean? Your husband and you or somebody else in the team? (55:28)

Sofya: We have a team of five. It is me, my husband who advises and helps me, and several former colleagues. Everyone else I found at meetups and they help with different parts of the work. (55:35)

Sofya: I cannot disclose all the names, but they work on coding, the application, design, interviewing people, and marketing. It cannot be done by two people alone so we needed a team in the beginning. I am thankful that I found people who are motivated to solve this problem and take on the challenge. (55:49)

Alexey: How did you find them? (56:28)

Sofya: Meetups mostly. I am a fan of going wherever life takes me. If I see an interesting event I go. Web Summit was a perfect example, though there are too many people to find someone properly there. (56:34)

Sofya: It is complicated to have a long conversation in that environment. I think we only talked for about five minutes before more people joined and we lost the thread. The entire day was like that. (56:52)

Sofya: Because of that I go to small meetups. I went to many in Latvia. In Latvia we had a place called Startup House, a co-working space only for startups. It is amazing because you share knowledge and chat over coffee about the challenges you are solving. (57:22)

Sofya: It was a great decision for me to move from an office to a co-working space because I met people around me. After hours there were many events like pancake breakfasts, presentations, pitches, marketing talks, and coding evenings. These events help you get to know people, share challenges, and understand who might be a good team member or who knows someone who could be. (57:55)

Alexey: Why did you move to London? Is it related to your company or did it just happen? (58:48)

Sofya: It is related because I see more opportunities here. The market is huge. (58:54)

Alexey: And I guess when it comes to finding investors London is active. My friends there tell me about the meetups for startup founders and people interested in startups. Every week there are several events. (59:00)

Alexey: Berlin is similar but the scale is smaller because the city is smaller than London. I see that we are already out of time. This conversation flew by and it was amazing, but we should wrap up. (59:22)

Alexey: Sofya, thanks a lot for joining us today. It was great talking to you. (59:34)

Closing insights on Sofya's educational path and early device prototypes

Alexey: Thanks for sharing everything with us. It was interesting to hear your story, your educational background, and how you created the first prototype and got to the device you have now. Thanks for joining us and telling your story. I wish you great success. (59:48)

Alexey: Maybe we will meet again at Web Summit if I go next year. (1:00:15)

Sofya: It would be nice to meet you. Thank you for inviting me. Thank you for a wonderful conversation. It was nice to hear and see you one more time. (1:00:22)

Alexey: Thanks everyone for joining us. Some of the questions I asked came from our attendees. Thanks for being active and asking questions. (1:00:33)

Alexey: That is it for today. Enjoy the rest of your week and the rest of this year. The holiday season is starting, though not for people who run companies because we keep working. For the rest of you, enjoy the holidays and see you soon. (1:00:46)