Translation. Region: Russian Federation –
Source: Peoples'Friendship University of Russia
An important disclaimer is at the bottom of this article.
One of the RUDN.VC accelerator projects is a chatbot for patients with type 1 diabetes. It was developed by Nikita Radaev and Dmitry Prutskikh, students in the Biomedicine program at the RUDN University Institute of Medicine.
The team created a multifunctional chatbot in Telegram that will take on disease management and solve problems:
Calculating all necessary insulin doses and carbohydrate (Bread Units) content in foods; keeping a self-monitoring diary; receiving reminders about procedures and refilling medications when they run low; calling an ambulance by pressing a panic button in the event of a sudden drop in blood sugar (hypoglycemia); and allowing parents to monitor their child with diabetes using a smartphone.
We spoke with the team and learned how Nikita's personal experience with illness led him to the idea for the project, how the chatbot functions, and what the developers are striving for.
Nikita, how did your personal experience living with diabetes help shape the idea for this project?
From the age of eight, I watched as first doctors, and then my grandmother and mother, performed countless calculations to determine the amount of carbohydrates in each food I ate so they could inject me with the right dose of insulin. Every child with type 1 diabetes has probably had this experience. I still remember my father drawing hand-drawn graphs showing changes in my blood sugar levels. It's impossible to forget the printed chart of carbohydrates per 100 grams of food, the mechanical alarm clock, and the thick school notebook with pen-lined margins that was destined to become a "Self-Monitoring Diary." Difficulties also arose due to the lack of reliable and up-to-date information on the disease—a significant problem at the time in Kyrgyzstan, where I was born and lived until adulthood. From the age of 12 or 13, I began to perform many of the calculations myself. This was tedious, as it required knowledge of numerous formulas and various coefficients—and any error was unforgivable. In ninth grade, I started learning Python programming, and I immediately knew which program would be my first. Within a few weeks, I had written a mini-app for my personal use that could solve two of my most pressing problems: calculating my insulin dose and reminding me to check my blood sugar every two hours after an injection.
How did the collaboration between you and Dmitry come about?
For a long time, I used the application I wrote only for personal needs. I can say that almost immediately my average blood sugar levels became much better. I continued to slowly but surely improve the code I wrote. Initially, there was no goal to bring such an application to the market, but realizing that it could really improve the lives of other people with diabetes, I decided to rewrite the program with cleaner and more user-friendly code and offer it to the public. I understood that this would require a lot of work, which I simply could not cope with alone, and decided to find support in the form of my classmates. Dmitry, one might say, has always been and remains an excellent student and the “brain” of our group, one of the most hardworking students, so I decided to approach him with such a proposal. Dmitry, despite his busy schedule and work in a scientific laboratory, agreed to help. Soon I learned that RUDN University was holding an accelerator of student projects RUDN.VC, and I decided to apply for participation in which. And to develop the project, our team needed an experienced, involved mentor. This was Natalya Vladimirovna Bolotova, whose acquaintance was a gift from the accelerator. I understood perfectly well that if the project was not at least minimally commercially successful, then we could forget about creating a truly high-quality and competitive product. Many, so to speak, strategic stages of the project, such as the presentation of our MVP, analysis of the target audience and the initial construction of a business model were completed only thanks to the help and knowledge in these areas of Natalya Vladimirovna.
What was the main catalyst that made you move from the idea of "it would be nice to have a bot like this" to real action and applying to the accelerator?
The catalyst was the realization that if my algorithm helped me, it could help others too. I have many friends and acquaintances with type 1 diabetes. They come from all over Russia and Kyrgyzstan, from neighboring countries and beyond, and have varying incomes and education levels. But almost all of them face the grueling routine of diabetes procedures and suffer from inaccurate calculations. If my chatbot helps users reduce the risk of complications from the disease, I will consider myself to have done a great and beneficial job for society.
Tell us about the "panic button" for calling an ambulance. How will this function be technically implemented?
The bot will recognize that a patient has lost consciousness and requires assistance if, while using the "Increase Low Blood Sugar or Relieve Hypoglycemia" mode, the user does not respond to messages within a specified time, for example, 10 minutes. In this case, the patient's geolocation and condition data will be transmitted to the nearest emergency department, and information will appear on the smartphone screen, allowing concerned passersby to help the user. The patient will also be able to activate the "panic button" using a dedicated function.
How will the medication purchase reminder function work?
The user will use the bot to record when they started using a new insulin cartridge in their pen, purchased test strips, or purchased needles. They will also track the amount of insulin administered and record blood sugar measurements. The bot will subtract the doses administered from the initial amount of insulin in the cartridge, accounting for two to three units of insulin expended during the pen's setup before each injection. When the bot calculates that the insulin in the cartridge is low, it will notify the user. A similar logic will be used for counting needles and test strips.
The bread unit (BU) database is a huge undertaking. How do you populate and verify it?
There are numerous proven nutrient density tables, compiled back in the 1990s and 2000s, but still relevant and widely used today. We decided to digitize them and use them as the basis for our bot. Naturally, we also include, so to speak, "new" products, such as avocado and lychee. nutrient density tables for packaged products are not as relevant, as manufacturers almost always list the protein, fat, and carbohydrate content per 100 grams of product themselves. Our bot provides a function that can calculate the nutrient density based on this information. Of course, it would be great if manufacturers themselves listed the nutrient density in their products, and our team would like to promote this idea to the public. I believe many manufacturers would welcome such an innovation, as it would only highlight their concern for people with special needs.
How is the medical validity of algorithms ensured, especially in terms of insulin dose calculations?
When calculating insulin doses, our bot analyzes a fairly large array of patient data and suggests a dose based on average values, so any "inflated" values are minimal. This approach is well-tested—in over three years of personal use, the program has never returned any extremely low or high insulin doses. However, the possibility of a bot malfunction cannot be completely ruled out. We've anticipated this scenario. The bot displays a notification if the insulin dose it calculates is outside the acceptable range.
What role does endocrinologist Anna Vetrova play in the project? Is she consulting during the development stage or will she be overseeing the medical content on a full-time basis?
Anna Vladislavovna is making, and has already made, a significant contribution to medically validating the bot's algorithms in accordance with modern diabetology methods adopted in Russia. We are currently working with her on the bot's FAQ feature. It will contain answers from a practicing endocrinologist and diabetologist to hundreds of frequently asked questions by patients with type 1 diabetes.
How does a medical education help in IT product development? And conversely, how has working on a startup influenced your perception of the future medical profession?
The curriculum for our specialty devotes many hours to natural science and research, rather than clinical, disciplines. This allows us to do what we study for—create new technologies and products that doctors, after preliminary review and approval, will then use in their clinical practice. Our startup is precisely that—a product from biomedical researchers for use by doctors and, through them, by patients. While developing the startup, we drew on many of the knowledge we gained at university, particularly in subjects such as normal and pathological physiology and biostatistics. Our ability to work with large data sets was also very helpful—to develop the algorithm, we had to read over 20 papers on diabetology. Working on the project helped us reaffirm our commitment to the right path and apply our acquired knowledge and skills in practice.
Has your project or its algorithms received any formal evaluation or approval from the professional medical community?
Yes, in early 2025, we presented our algorithm code to the National Center for Maternal and Child Health (NCMCH) and the City Endocrinology Dispensary in Bishkek, Kyrgyzstan, where our work was highly praised and approved. We received a letter of recommendation from the NCMCH management confirming their willingness to use the bot as a supplement to the so-called "Diabetes School"—lectures where a doctor teaches newly diagnosed patients about living with diabetes. Once the bot is ready, we plan to share it with clinics in Russia for recommendations and approval.
Do you plan to monetize the project?
Although profit isn't the project's primary goal, we'll still need to implement monetization to maintain its viability and further development. There will likely be a free ad-supported version of the app and a paid version with expanded ad-free functionality. The estimated subscription price for this year is between 100 and 250 rubles per month. We're also considering adding a voluntary donation system for project development.
Your ambition is to expand beyond Telegram. What platform is your next priority (for example, a dedicated mobile app) and why?
To begin with, we decided to implement our project as a chatbot rather than a standalone app for a number of objective reasons: the simplicity of writing chatbot code, the ease of testing hypotheses, and the ability to quickly improve functionality and fix bugs. Once our algorithm reaches a sufficient level of quality, we will consider creating a standalone app for popular operating systems. The main advantage of using an app rather than a chatbot will be the ability to run the algorithm's core functions without an internet connection. At the same time, we haven't abandoned the idea of further developing our Telegram bot. In fact, we are considering creating similar bots for other messaging apps, such as the national messenger MAX, VK, and even WeChat, which is popular in China and many other countries. In the long term, we also plan to add new languages to the original Russian and English: Spanish, French, and Portuguese, which are common in Russia-friendly countries of South America and Africa, as well as Chinese, Arabic, Mongolian, and many languages spoken in the CIS countries, such as Kyrgyz and Kazakh. Our mission is to expand the project as widely as possible globally, enabling millions of people to improve their diabetes self-management for free or at a small cost.
Please note: This information is raw content obtained directly from the source. It represents an accurate account of the source's assertions and does not necessarily reflect the position of MIL-OSI or its clients.
