Center of Trustworthy AI for Life Sciences is an International Research Agenda project at the University of Warsaw. The Center develops trustworthy, efficient and interpretable AI methods for medicine and the life sciences. We build new algorithms for analysing multimodal data, including EEG signals, clinical descriptions, text, images and other biological measurements. Our goal is to create AI systems that do not only provide predictions, but also explain their decisions, quantify uncertainty and remain reliable when data are noisy, incomplete or limited.

A central focus of the Center is the automated analysis of EEG signals. EEG is one of the most important clinical tools for studying brain function and diagnosing neurological and neurodegenerative disorders, but it is still difficult to analyse at scale because the signal is complex, noisy and highly variable. We will develop and validate new AI technologies using access to one of the world’s largest EEG resources, containing over 250,000 clinical recordings and more than 10,000 repeated examinations of individual patients, together with medical descriptions and diagnostic information.

The programme is built around four closely connected research groups.

Artificial Intelligence for Human Biology lead by dr hab. Jacek Rogala develops and validates AI methods for biological and clinical data, with a special focus on EEG. The group connects algorithmic development with medical interpretation and clinical validation, ensuring that the technologies created in the programme address real diagnostic needs.

Effectiveness of Foundation Models, led by Prof. Błażej Miasojedow develops efficient AI systems and foundation models for multimodal data. The group focuses on scalable architectures, mixture-of-experts models, distributed training, model distillation and fast inference, so that advanced AI can be used in practical and cost-effective applications.

Mathematical Foundation of AI, led by Prof. Anna Gambin, develops mathematically grounded methods for reliable and explainable prediction. The group works on feature selection, uncertainty quantification, optimal transport and methods that allow AI systems to assess the confidence and risk of their own outputs.

Topological Features of Data, led by Dr. hab. Paweł Dłotko, develops topological, geometric and statistical methods for extracting meaningful structure from complex data. The group uses topological data analysis to build informative features for AI systems and to better understand how models make decisions, when they fail and how they can be made safer.

What makes this agenda distinctive is the close integration of advanced mathematics, modern AI and real biomedical data. The four groups work on a shared technological pipeline: from data quality control and feature extraction, through efficient model construction, to uncertainty analysis, explainability and clinical validation. This structure allows the programme to address problems where standard AI often struggles: small datasets, heterogeneous clinical sources, missing data, domain shifts and high levels of noise.

The long-term ambition of Center of Trustworthy Artificial Intelligence for Life Sciences is to create AI technologies that can support earlier detection of disease, improve the safety and reliability of medical AI, and open new routes for applying mathematics, statistics and topological data analysis to real-world biomedical challenges.

Center of Trustworthy Artificial Intelligence for Life Sciences gratefully acknowledge support from the Project FENG.02.01-IP.05-M009/25 implemented as part of the International Research Agenda Programme of the Foundation for Polish Science, financed by the European Union from Priority 2 of the European Funds for a Modern Economy 2021–2027 Programme (FENG)