Dario Coscia

prof_pic.jpg

Hi I’m Dario! I’m a PhD student in the MathLab group at the International School for Advanced Studies and at the University of Amsterdam, supervised by Prof. Gianluigi Rozza and Prof. Max Welling. My research interests focus on the application of deep learning for solving differential equations. Specifically, I am driven to investigate deep learning methods for unstructured data, unsupervised generative learning for uncertainty quantification, and physical constraints for ML models. If you are interested in the topics, check out also the medium series Deep Learning 4 Natural Science.

Previously, I was a master student at the University of Trieste in Data Science and Scientific Computing, where I graduated with a thesis on Generative Modelling for fluid simulations. My Bachelor was in Physics, where I conducted one research internship at CNR-IOM on simulating the energetics of graphene on different material sources.

news

Jun 26, 2025 I gave a talk at Banff International Research Station during the workshop Efficient and Reliable Deep Learning Methods and their Scientific Applications, talk available here.
Jun 18, 2025 I gave two lectures at the KTH Royal Institute of Technology during the PINN Summer School: PINA software (slides and code) and Uncertainty Quantification in Deep Learning (slides and code).
May 12, 2025 I lectured in the Advanced School on Foundation Models for Scientific Discovery at the International Centre of Theoretical Physics: Introduction to Model Uncertainties in Deep Neural Networks, slides; Uncertainty in Scientific Machine Learning from PDEs to Molecules, slides.
May 12, 2025 I partecipated in the Round Table Discussion on Uncertainty Evaluation during the Summer School on Theory, Mechanisms and Hierarchical Modelling of Climate Dynamics: Artificial Intelligence and Climate Modelling at the International Centre of Theoretical Physics.
Apr 29, 2025 I gave a talk at Heidelberg University’s Institute for Theoretical Physics on Uncertainty Quantification for Sequence Models slides here.