Tomasz Kuśmierczyk: About me
I am a machine learning researcher working on probabilistic modeling, Bayesian deep learning, and reliable AI systems. I am interested in bridging statistical inference and modern deep learning to develop models that are more robust, interpretable, and effective in real-world settings.
My recent work is on uncertainty quantification for large language models: making Bayesian treatment of large models affordable through parameter-efficient adapters (Findings of EMNLP 2025), and distilling expensive uncertainty-aware teachers into models that need a single forward pass (ECML-PKDD 2026). From 2024 to 2026 I led the NCN/MSCA project Balancing priors and learning biases to improve Bayesian Neural Networks at the Jagiellonian University, as principal investigator. Earlier I was a postdoctoral researcher at the University of Helsinki, and I hold a PhD from NTNU Trondheim.
GitHub
LinkedIn
Scholar
DBLP
2024-2026 Project
ResearchGate
News (Twitter)Research activity
- Funding
- Principal investigator of Balancing priors and learning biases to improve Bayesian Neural Networks, 2024–2026, at the Jagiellonian University. Funded by the National Science Centre under POLONEZ BIS 2 and co-funded by the European Union Horizon 2020 Marie Skłodowska-Curie programme.
- Reviewing
- Program committees: AAAI 2026 and 2027 · UAI 2026 · ECML-PKDD 2026 · ICML 2024 · NeurIPS 2024. Also reviewing for AISTATS and the journal Knowledge and Information Systems.
- Teaching
- Guest lectures on Introduction to Bayesian Methods and Bayesian Neural Networks at the Jagiellonian University, 2024 (long version, short version). Nordic Probabilistic AI School (ProbAI): co-organizer 2019 in Trondheim and 2022 in Helsinki, head teaching assistant 2022, teaching assistant 2024 and 2026.
- Supervision
- Doctoral researcher on the PLBNN project, 2024–2026. Co-supervised one master's and one doctoral student at the University of Helsinki, 2018–2019.
- Invited talks
- A Bunch of Tricks for Bayesian Neural Networks, Multi-source Probabilistic Inference group, University of Helsinki, 2026 · Challenges in Bayesian Neural Networks, Centre for Credible AI seminar, Warsaw, 2026 (slides) · High-Fidelity Transfer of Functional Priors for Wide Bayesian Neural Networks, Institute of Computer Science, Polish Academy of Sciences, 2024 (slides) · GMUM Tea Seminars, Kraków, 2024–2025 (slides)