ITMO Students Secure 5.8 Million Rubles at International AI Challenge

Several ITMO students got the top spots in the teams track. The first place and 1.5 million rubles for solving a problem suggested by SberAI went to the team 12,5 Squad that included Alexander Smirnov, a student of ITMO’s Artificial Intelligence Technologies Faculty. Together with his teammate Alexey Shaturin, he developed a tool for efficient search in three modalities – texts, images, and 3D objects.

“We trained the model to find identical objects in different representations (photos, texts, or 3D maps). For each representation, we used specialized neural networks called encoders. They convert data into numerical vectors containing information about the encoded objects. Using contrastive learning, the encoders were trained so that vectors of the same object from different modalities would be similar, while vectors of different objects would differ. The main advantage of our solution was a powerful 3D encoder that works with CAD objects as both images and point clouds simultaneously. Thanks to AI Challenge, I discovered another field in ML connected to CAD objects processing. I was surprised by the diversity and originality of its approaches and architectures, and in the future I may continue studying this field,” says Alexander Smirnov. 

The second place for solving a case from SberAI went to the team Oatmeal Cookies Ambassadors from ITMO’s Faculty of Software Engineering and Computer Systems. Alexey Korobov and Timur Sarvarov won 900,000 rubles.

Two more teams fom ITMO solved a problem from Sber’s Health Industry Center. The first place went to team MISS x alpha beta gamma stroke, which included Timur Ionov, a PhD student at ITMO’s Institute of Applied Computer Science. For developing an efficient tool for analyzing suggestions offered by a digital pharmacist to human pharmacy operators and improving the accuracy of its recommendations, the team received 1.5 million rubles.

The third place and 600,00 rubles went to Egor Andreasyan, a first-year at the Artificial Intelligence Technologies Faculty, who was part of the Wassup! team. According to Egor Andreasyan, it’s not his first AI Challenge. He believes that this invaluable experience of working on cases for major companies will help him become a qualified specialist capable of solving complex tasks in AI.

In the individual track, Diniyar Murtashin, a first-year at ITMO’s Artificial Intelligence Technologies Faculty, became the winner in the category School students. At the moment of registration for the competition, he was studying at a school in Sterlitamak and later enrolled at ITMO thanks to his performance in the National Technological Competition in big data and machine learning. At AI Challenge, Diniyar presented the best solution for a problem by the Artificial Intelligence Research Institute (AIRI): he trained a model to predict the probability distribution of white points for an image, which will help perform color correction. For his win, Diniyar received 500,000 rubles.

“The hardest part for me was reaching a bitwise reproducibility of the solution. The rules in this regard were very strict: if the metrics weren’t reproducible, participants were banned even if they had the perfect presentation and a top position on the leaderboard. There were also a bunch of criteria to meet in terms of project defense, and I needed to analyze the model’s work in detail. All in all, it took me several days to collect all the data and make the necessary plots,” shares Diniyar Murtashin. 

Daniil Antonov, a first-year student of ITMO’s Software Engineering and Programming Faculty, took the third place for developing an AI agent for Rosatom. Together with his teammate Artem Mazur (team Sota), he developed an agent capable of optimizing the energy state of a graph using the Ising physical model. For their solution, students received 400,000 rubles.

Team The Boys and Konstantin Sigalov, a first-year student of ITMO’s Software Engineering and Programming Faculty, took the third place and was awarded 400,000 rubles for solving a case by the Central University. The team developed a system for solving mathematical problems and a model capable of accurately interpreting the problem and providing the correct answer in a numerical or symbolic form.

Results of another international AI competition by Sberbank and AIRI, AI Journey Contest, were announced at the eponymous conference. 1,500 contestants from around the world competed for a 6.5 million-ruble prize fund, among them was ITMO graduate Ammar Ali. He won two of the contest’s three tracks, having developed the best judge model to evaluate the generative capabilities of LLM agents in the track Agent-as-Judge, as well as long-term memory for a language model in the track GigaMemory: Global Memory for LLM – with his teammate Valaa Saman. His two wins brought Ammar 1.55 million rubles.

This became Ammar Ali’s fifth win at AI Journey Contest – and one of many in his outstanding portfolio. For instance, as an ITMO student, he won the Image Matching Challenge by Google Research twice. In an interview with ITMO NEWS, Ammar shared his preparation tips for machine learning competitions, as well as the reasons he finds contests in the field so appealing.

AI Challenge is an international competition in AI for youth, organized by Sberbank and the Artificial Intelligence Alliance. This year, 10,000 school and university students from 75 countries around the world took part in the competition. The winners were 12 teams in the categories University students and School students, as well as 18 participants in the individual track in the categories Beginner, School students, and University students. The total prize fund of AI Challenge was 15.6 million rubles.