TuneAI supports various tasks. Lecturers can use it to test students’ knowledge, namely to create tests and adjust grades manually if needed, students may find it useful for self-checks before an exam or a test, and companies, in turn, can upload their questions to interview candidates.
“We know firsthand that preparing for exams and job interviews can be a challenge. You need to gather a ton of materials and sort through all of them, but it’s still hard to assess yourself honeslty. Some choose cramming; others revise theory in fragments – either way, when you try to test yourself, it’s not easy to tell what you genuinely understand and what you don’t. In this case, some people opt for ChatGPT; they upload their materials there and ask the system to test them. However, the problem is that they still have to set the grading criteria and feedback format manually in the chat. Hence our solution, TuneAI – it can be tailored to your goals and provides detailed feedback,” says Maksim Chusovlyanov, a student of the System and Applied Software Bachelor’s program and a lead developer of TuneAI.
Grigory Sadovoy and Maksim Chusovlyanov. Credit: Yandex
The platform works as a custom builder. Users create their own libraries by uploading lectures, study guides, and presentations for the system to use during examinations. Materials can be linked at different levels: organization, course, test, or individual question. Users can set custom assessment criteria: which answers are correct, how many points are assigned for each, and how strict the assessment should be. Tests can include different types of questions (open, short answer, and single- and multiple-choice) and individual formats for each of them. Users can submit their answers via text or voice messages.
“User responses undergo multiple processing stages. Voice messages first need to be converted into text – it’s done by SpeechKit, a speech recognition service. Next, the built-in YandexGPT receives the answer, the question, the evaluation criteria, and relevant excerpts from the learning material. Based on this data, the system identifies what was said correctly, what was missed, and where the user made a mistake. To search for the revelant excerpts, we use text embeddings – these are vector representations of text that help the system filter the uploaded materials to find those relevant to a specific question. As a result, students understand what they know and what they need to review, lecturers spot answers that need attention, and interviewers can find the best-fit candidates,” shares Grigory Sadovoy, a student of the System and Applied Software Bachelor’s program and an author of the TuneAI project.
Grigory Sadovoy and Maksim Chusovlyanov at YaC/e 2026. Credit: Yandex
Such platforms aren’t yet widespread in Russia. However, there are tools with similar features on the global market. For example, in Google Classroom, users can upload their lectures and presentations into the system and take tests based on their uploads.
The product is available as a demo website that opens on any device, including smartphones. The website is adapted for mobile browsers, and its source code can be found on GitHub. The developers plan to integrate their tool into Moodle, a popular course management platform used, among others, at ITMO. Right now, the team is finalizing its role-based model: a student, an examiner, a candidate, a methodologist, a teacher, an interviewer, and an administrator. It’s expected that some roles will be able to take tests and answer questions, while others – create tests, assign them to users, and review controversial answers manually. Among other features to appear is assigning a test to an entire group and hiding further questions until the examinee answers the previous ones.
Grigory and Maksim worked on their tool in their open-source ITMO student club CRUDoshleps. Their club’s name is a pun on the CRUD acronym, which stands for the four basic operations with databases: create, read, update, and delete.
The platform was presented at the poster session at YaC/e-2026, one of Russia’s top conferences focused on technologies and new trends in education. By presenting their product, the team was able to showcase their advances and attract new members. In the future, the developers will use their experience to gain official status – for this, they will need a complete and pitched project.
