A more formal version is available here: click to open or directly download
A brief overview of my reserch interests and current projects
My research lies at the intersection of data assimilation, uncertainty quantification, and machine learning, with applications focusing on geophysical systems. My current project involves deep learning based end-to-end trainable models for ocean data assimilation and forecasting, as part of the Horizon Europe EDITO Model Lab, which focuses on the development of the European digital twin ocean. I am developing ensemble generation in end-to-end neural DA schemes in 4DVarNet, a novel neural DA scheme, 4DVarNet-LU, to improve both accuracy and uncertainty quantification at inference.
Earlier during my PhD, I worked data assimilation for chaotic dynamical systems using ensemble Kalman filters, where I demonstrated numerical filter stability, a crucial property of a filter, using optimal transport-based distance between probability distributions and stability of computing Lyapunov vectors from erroneous trajectories. I am particulary fascinated by problems that arise in earth sciences from a dynamical systems perspective. Data assimilation is a crucial part that makes numerical weather prediction possible. I work on both classical data assimilation and deeplearning based data assimilation techniques, most of the times looking for an ideal way to use the best of both worlds.
Future Research Objectives
Academically trained as a physicist, problems at the interface of climate science, data, algorithms and society are something that inspires my long-term vision. My understanding of both physical and algorithmic aspects enables me to approach complex scientific challenges. Broadly, I would like to work on challenges focusing on future weather prediction and climate models to model and capture extreme events at both regional and global scales. This requires understanding the key drivers of physics-based models and their contribution in the control/ correction of data-driven models, in terms of accuracy, uncertainty and risk. Developing methods to handle and model uncertainty, dynamical knowledge, combined with general machine learning techniques for working with big datasets, is what I would like to develop my skills further in. In the following pages, I briefly describe my current and previous research projects of interest.
