Short Bio. I am with the Instituto de Telecomunicações, Instituto Superior Técnico. I obtained my BSc. and MSc. from Instituto Superior Técnico (IST), Lisbon-Portugal, and Ph.D from Carnegie Mellon University, Pittsburgh-USA, and IST all in Electrical and Computer Engineering. I held a 2-year Postdoctoral scholar position at Carnegie Mellon University and another 2-year at the École Polytechnique Fédérale de Lausanne (EPFL), Lausanne-Switzerland. 

Research Interests. I am mostly interested in the qualitative analysis and statistical inference of complex systems. In particular, in developing interpretable and explainable ML or AI strategies to unveil causal structure from data within the scope of high-dimensional Networked Dynamical Systems. The problem refers to recovering the fundamental dependencies between the nodes/variables/components comprising a complex interacting system from observation of the activity (or time series samples) stemming from the nodes. Examples include Brain activity, pandemics and many more. In these applications, the time series reflect the state evolution of the dynamical system and are readily available, e.g., in the form of EEG or FMRI signals from distinct regions of the Brain or from the report of the number of infected individuals across distinct communities in a pandemics. However, the underlying core causal structure (the network!) lies hidden and it is the goal of causal inference to unveil consistently this connectivity pattern from the time series. Further, in the high-dimensional setting (which comprises most applications), it is unfeasible to probe the activity over all nodes in the system and it is crucial to develop methods tailored to partially observed systems (with technical guarantees of structural consistency).  

Recent Contributions. i) We have recently proposed a novel ML causal inference paradigm with full explainability (i.e., guarantees of structural consistency) for certain large-scale linear networked dynamical systems -- refer to the AAAI publication in Selected Publications or to my recent online presentation at underline following the link provided below; ii) We proved that the Granger estimator is structurally consistent (i.e., faithfully recovers the underlying causal structure from time series) for certain large-scale linear stochastic networked dynamical systems under partial observability (see references 2,3 and 4 in Selected Publications); iii) We have shown that certain powers of the precision matrix (computed from FMRI time series) convey non-trivial information about the underlying connectome structure in the Brain (see reference 5 in Selected Publications). 

Acknowledgement. Our group's main goal is to devise a principled ML/AI-based platform for structure learning high dimensional networked dynamical systems under distinct dynamical laws and broad regimes of observability, network connectivity and noise structure. The project is partially supported by the U.S. National Science Foundation (NSF) under the Grant CCF-2327905.