Jorge FERNANDEZ-DE-COSSIO-DIAZ

I am a researcher interested in statistical physics, computational biology, and machine learning.

My interests lie at the interface between the statistical physics of learning and computational biology. I develop generative models and representation-learning algorithms, try to understand their behaviour with tools from statistical physics, and apply them to biological systems, such as proteins or RNA, often in close collaboration with experimental teams.

I lead a group at IPhT and regularly have open positions for interns, PhD students and postdocs. I also teach statistical physics and machine learning at École Polytechnique and in Paris-area master programs.

news

Oct 11, 2026	I’ll be lecturing at the IHP School on Artificial Intelligence for Mathematicians and Theoretical Physicists, Institut Henri Poincaré, Paris, 22–26 February 2027.
Oct 11, 2026	I’ll be giving a talk at the Forum de la théorie au CEA, 9–10 December 2026.
May 14, 2026	New preprints: Spherical Boltzmann machines (NeurIPS 2026, Paris, and HiLD workshop, ICML 2026), Replica theory of spherical Boltzmann machine ensembles, and Steering protein language models with Iterative Lookback Monte Carlo (spotlight at the GenBio workshop, ICML 2026).
May 14, 2026	Our paper on shared latent representations of zebrafish brain activity across individuals is published in PNAS. See also the CNRS Physics commentary.
Jan 12, 2026	New preprints: cross-individual translation of zebrafish brain activity, specificity-switching mutational paths of WW domains (now an eLife reviewed preprint), and a high-order cumulant extension of quasi-linkage equilibrium.
Dec 18, 2025	Our paper Designing molecular RNA switches with Restricted Boltzmann machines is out in Nature Communications. CNRS Physics commentary: English, French.