Med AI Lab
Innopolis University, Russia
We work on medical AI where the clinician is part of the system, not a user of it. Our most developed line is gaze analysis: what a radiologist’s gaze says about the quality of their work, and how that signal can be used to train models. Alongside it we run work on electronic health records and on the assessment of IVF embryo images. We work closely with clinicians and with private datasets, run our own eye-tracking experiments with practising radiologists, and work directly with leading clinical centres.
People
Ilya Pershin · Head of Laboratory
Clinician gaze, medical imaging, human attention in language models
Dmitry Lvov · Researcher · PhD student, ITMO University
Electronic health records
Ivan Stebakov · Researcher · PhD student, Orel State University
IVF embryo assessment, histology scans
Daria Galushko · Researcher · BSc student, Innopolis University
Eye tracking for non-invasive risk profiling of neurodegenerative disease
Maxim Izyumov · Researcher · BSc student, Innopolis University
Electronic health records
Artem Grishin · Researcher · BSc student, Innopolis University
Radiologist gaze data
Arina Petukhova · Researcher · BSc student, Innopolis University
Electronic health records
Kirill Beukh · Researcher · BSc student, Innopolis University
Electronic health records
Zhanna Ivanova · Researcher · BSc student, Innopolis University
IVF embryo assessment
Alumni
Leila Khaertdinova · now at the University of Copenhagen
Eye tracking of radiologists
Karim Galliamov · now at the University of Amsterdam
RLHF from eye-tracking data
Tatyana Shmykova · now at Innopolis University
Gaze-based volumetric segmentation
Clinical partners
- Kulakov National Medical Research Center for Obstetrics, Gynecology and Perinatology
- Loginov Moscow Clinical Scientific Center
- Pirogov Russian National Research Medical University
Collaborators
- Ivan Titov — University of Edinburgh and University of Amsterdam. Three joint papers: EMNLP 2023 (main track), EMNLP 2025 (main track), ICLR 2026 workshop. The lab’s work on AI interpretability and post-training methods originated with him: joint work on gaze-based RLHF and concept bottleneck models.
- Bulat Ibragimov — University of Copenhagen. Co-author on both IEEE Access Q1 papers and on the radiologist gaze line from 2022 onwards. The lab’s eye-tracking research originated with him.
- Tamerlan Mustafaev — University of Pittsburgh. A radiologist who shaped the clinical framing of our tasks and was the first participant in the lab’s eye-tracking experiments. Co-author on the radiologist gaze line.
Joining the lab
What we work on. Eye tracking of clinicians, medical imaging, electronic health records and the modelling of human attention in language models. More on the research page.
What you get. Co-authorship at A* venues. In 2025–2026 the lab produced 13 such papers, two of them in main track (EMNLP, NeurIPS), and two Q1 journal papers. Several of our undergraduates are co-authors on the work listed under publications.
Where people go afterwards. Recent lab members have gone on to the University of Copenhagen and the University of Amsterdam.
What we look for. Python, a working knowledge of machine learning, analytical thinking, and enough English to read papers and to write them.
Funding. Paid positions are possible and are discussed individually.
How to apply. Write to i.pershin@innopolis.ru with a short account of what you have already built, what you would like to work on, and a research proposal.