Short bio

My master thesis was devoted to the algorithmic side of Network Science, i.e. pattern matching and counting, specifically on the subgraph isomorphism problem.

I then decided to give industry a try and was a data scientist for about a year and half. This was an exciting time as Transformers had just made their debut so I had the opportunity to learn and apply these architectures first hand. Their mysterious performance also made me realize where I stood in terms of the Dunning-Kruger effect and motivated me to double down on my studies.

Since 2020 I started my PhD. in Computer Science at the University of Porto. The main jist of my thesis centers around inductive priors that capture our empirical knowledge of some phenomena with data-driven models, very much in the line of Model-based deep learning. Examples span from Markov chains that model the regular ticking of the heart, to the fractal character of pathological phenomena as observed in several Medical Imaging modalities.

Currently I have set my gaze in self-supervised learning. Everyday I become more convinced that reality can be codified with few parameters. But how few? At what cost?