CV
A short overview of my background, talks, teaching and service. The full CV is available as a PDF via the icon on the right.
Contact Information
| Name | Matthias Kellner PhD student in computational materials science |
| matthias [dot] kellner [at] epfl [dot] ch | |
| Website | https://kellner.science |
Professional Summary
I develop machine-learning methods for atomistic simulation, with a focus on uncertainty quantification. I am interested in determining when a model’s predictions can be trusted and when they cannot, which is a question central to deploying ML reliably in the physical sciences. I am also involved in the development of universal machine learning potentials. I have developed the universal PET-MOLS model and applied it to study amorphously formulated active pharmaceutical ingredients. I also build deep-learning models for predicting NMR chemical shieldings (ShiftML), which I make available to the wider chemistry community.
Experience
-
2023 - present Lausanne, Switzerland
Doctoral Researcher
EPFL Lausanne, labCOSMO (Group of Prof. Michele Ceriotti)
- Develop uncertainty-quantification methods for machine-learning interatomic potentials.
- Build deep-learning models for NMR chemical shieldings in molecular organic solids (ShiftML).
- Develop universal machine learning potentials for condensed organic matter (PET-MOLS).
-
2019 - 2021 Darmstadt, Germany
Working Student
TU Darmstadt, Krewald group, in collaboration with Janssen Pharmaceutica
- Computational study of stereoselective C-glycosylation reactions.
Education
-
2023 - present Lausanne, Switzerland
PhD
EPFL Lausanne, labCOSMO
Materials Science
- Advisor: Prof. Michele Ceriotti.
- Expected graduation in 2027.
-
2020 - 2022 Munich, Germany
M.Sc.
Technical University of Munich
Chemistry
- Master’s thesis: Exploring multifidelity machine learning for chemical shift predictions.
-
2016 - 2020 Darmstadt, Germany
B.Sc.
Technical University of Darmstadt
Chemistry
- Bachelor’s thesis: Following chemical reactions using dDNP NMR spectroscopy.
Invited Talks
- PASC 2026 - Minisymposium, Trustworthy MLIPs - End-to-End Uncertainty Quantification for Atomistic Machine Learning. Bern, 2026.
- CAMD DTU seminar - End-to-End Uncertainty Quantification for Atomistic Machine Learning. Copenhagen, 2026.
- CCP-NC Seminar - Overview of ShiftML, QNC-NMR and open questions. Online, 2026.
- XXVI Swiss NMR Symposium - Quantum-corrected NMR crystallography at scale. Lausanne, 2026.
- CECAM Workshop on Uncertainty Quantification in Atomistic Modeling - Practical uncertainty quantification for atomistic simulations with last-layer ensembles. Lausanne, 2025.
- Machine Learning for NMR Crystallography Workshop, CCP-NC - Advancing chemical shielding predictions in organic solids. Manchester, 2025.
- SIAM Conference on Mathematical Aspects of Materials Science - Uncertainty quantification by direct propagation of shallow ensembles. Pittsburgh, 2024.
Contributed Talks
- DPG Spring Meeting - NMR crystallography at finite temperatures. Dresden, 2026.
- NMRlipids Database Hackathon - Chemical shielding predictions in organic solids and beyond. Bergen, 2025.
- DPG Spring Meeting - Advancing chemical shielding predictions in organic solids. Regensburg, 2025.
- DPG Spring Meeting - Uncertainty quantification by direct propagation of shallow ensembles. Berlin, 2024.
- Workshop on Uncertainty Quantification in Molecular Simulation, MPI Magdeburg - Uncertainty quantification by direct propagation of shallow ensembles. 2024.
Teaching
- EPFL - MSE-211 Organic Chemistry Lab (2023, 2024, 2025), lab instructor (small-group teaching).
- EPFL - MSE-305 Introduction to Atomic-Scale Modeling (2024), teaching assistant.
- EPFL - MSE-421 Statistical Thermodynamics (2025), teaching assistant.
- ICTP-MARVEL summer school (2026), tutor.
- TUM - Python programming for chemistry students (2022), tutor.
Outreach, Service and Reviewing
- Industrial outreach - Merck Innovation Cup, 2022.
- Student outreach - Journée des gymnasiens, EPFL, 2025.
- Reviewing - JCTC (2024), Nature Communications (2026), JCP (2026).