About Me
Hi, I am Giovanni 👋! I am a Postdoctoral researcher at the People-Centered Computing Lab at the Faculty of Informatics of the Università della Svizzera Italiana (Lugano, CH), led by Prof. Silvia Santini and Prof. Marc Langheinrich.
Previously, I have obtained my Ph.D. in Computer Science from the University of Liverpool (UK), under the supervision of Y. Goulermas, V. Gusev, M. Gaultois, and M. Rosseinsky. I received my Master's degree in experimental particle physics from the University of Pisa (IT), conducting my thesis within the Mu2e experiment at the Fermi National Laboratory (US).
My Research
My research focuses on interpretability-by-design: building deep learning models whose reasoning is interpretable by construction, which I study across model design, theory, and open-source code.
Among data modalities, I have a particular interest in time series and, more specifically, in what interpretability can and should mean for deep models that reason over temporal data.
You can find a list of my publications here.
Fundamental Research
Interpretable AI
I design deep learning models that are interpretable by construction, rather than explained post-hoc. This makes them steerable and verifiable.
Time Series
I develop deep learning for time series and spatio-temporal data: forecasting, classification, representation learning, and virtual sensing at unmonitored locations.
Applications
Physiological Signals
I apply my ML research to physiological and medical signals, used to infer stress, cognitive load, and engagement.
Natural Sciences
I also explore applications in the natural sciences, including experimental physics, climate, and materials science.
Code
PyTorch Concepts
Nov 2025 — ongoing PyC is a Python library built upon 
PyTorch and 
PyTorch Lightning to fuel research and facilitate benchmarks in interpretable deep learning.
Industry Projects
Weathering Predictions
Nov 2020 — ongoingWe use full-history weathering data for paints formulation to predict long term performances in untested locations. Ultimately, the extraction of greater formulatory information can guide the design of safe and durable materials.
Funded by Beckers