My current research focuses on developing a continuous-time framework for the system identification of stable linear systems. This work is deeply motivated by systems neuroscience, with the ultimate goal of building control-theoretic models of how biological brains perform model identification in real time.

Traditional system identification frameworks typically rely on discrete data sampling and the offline construction of large data matrices: processes that are fundamentally incompatible with the continuous, streaming nature of biological neural computation. My work instead pursues identification schemes that operate entirely in continuous time, learning a system's structure and parameters online as signals arrive.

At the Adaptive Systems Laboratory (EPFL), I worked on multi-agent reinforcement learning in decentralized partially observable Markov decision processes (Dec-POMDPs). This included developing methods for policy evaluation under partial observability, studying belief sharing mechanisms between agents, and analyzing how communication structure affects learning in decentralized settings.