Deep-Learning emulators:
The main focus of my research is the development of deep learning emulators for atmospheric physics. In particular, I am currently developing an emulator for a Radiative Transfer Model capable of simulating the full radiance measured by the European Space Agency's FORUM mission. My research interests also include the design of latent-space representations that effectively capture the underlying physical processes, as well as the development of physics-informed loss functions that better approximate the governing physical problem. More broadly, I am interested in artificial intelligence methods tailored to scientific applications, including attention mechanisms, interpretable models, and techniques that improve the physical consistency and generalization of neural networks.
Data Assimilation:
Data assimilation is a fundamental component of modern Numerical Weather Prediction, as it combines observations with model forecasts to produce accurate initial conditions. In this context, my research focuses on the development of deep learning-based observation operators, which offer a promising alternative to traditional approaches by improving the representation of the observation–model relationship.
