Bálint Máté

Pronounce á as in spa and é as in café, so my surname has two syllables, unlike the English word mate.
I work at Ångström AI, applying machine learning to improve computational techniques in chemistry and materials science. I live in a small town close to Zürich, Switzerland.

I obtained my phd in computer science and physics focusing on generative modeling, sampling and free-energy estimation. My supervisor was François Fleuret. I also had the chance to intern with the chemistry team of FAIR at Meta working on organic crystal structure prediction; to visit the group of Tristan Bereau where I learnt about free energies; and to spend some time in the electronic structure team of Microsoft Research where I played a small part in their effort to machine learn the exchange-correlation functional.

Before all this, I studied theoretical physics and differential geometry in Hamburg and mechanical engineering in Budapest.

[scholar] [github] [twitter] [email]

Selected Publications

See full publication list on Google Scholar →

Solvation Free Energies from Neural Thermodynamic Integration

2025

The Journal of Chemical Physics 162 (12), Editor's Pick

B. Máté, F. Fleuret and T. Bereau

[pdf] [summary]

Neural Thermodynamic Integration: Free Energies from Energy-based Diffusion Models

2024

The Journal of Physical Chemistry Letters 15 (45)

B. Máté, F. Fleuret and T. Bereau