I am interested in Bayesian statistics and machine learning, having focused on Bayesian optimization with Gaussian process surrogates and amortized likelihood-free inference.

For instance, while there has been a push to develop more black-box methods, bespoke approaches can exploit additional information to increase sample efficiency. As part of my PhD research, I have designed and developed a class of informative priors over functions that leverage nonstationarity to encode preferences, accelerating optimization and inference even under weak prior information.

I have also proposed a new active sampling method that leads to more informative samples compared to the traditional acquisition methods used in Bayesian optimization. This method is based on the concept of sequential Laplace approximation, or sequential mode-seeking variational inference, with mode collapse.

Education

MSc in Artificial Intelligence, University of Edinburgh, best student (Howe Masters Prize), 2018.
Specialization: Bayesian Statistics and Machine Learning.

MSc in Electrical & Computer Engineering, Instituto Superior Técnico, University of Lisbon, top 1%, 2015.
Specialization: Signal Processing, Telecommunications and Embedded Systems.