Fabrizio Pittorino is an Assistant Professor at the Department of Electronics, Information and Bioengineering (DEIB), Politecnico di Milano, working at the intersection of deep learning theory, adaptive AI, and efficient machine learning. He received a PhD in Physics (with honors) and later completed an MPhil in Statistics, developing a rigorously interdisciplinary research profile that bridges statistical physics, probability, and modern machine learning.

His main scientific goal is to build theoretical and algorithmic foundations for ML systems that must operate in real operating conditions: limited compute and memory (embedded/edge and integer-only pipelines), and non-stationary data (distribution shift, concept drift, continual adaptation). He focuses on the design principles needed for models that are simultaneously efficient (e.g., compressed, low-precision, hardware-aware, suitable for embedded/edge devices) and adaptive (capable of tracking distribution shifts and concept drift, and continually updating without losing robustness). A unifying scientific theme in his work is understanding how the geometry of loss and solution spaces governs optimization and generalization in deep neural networks, and how these insights can guide principled choices across algorithms, architectures, and representations. He designs theory-guided methods for efficient, dynamic and adaptive architectures, including flatness-aware neural architecture search, binary neural networks, and information-theoretic and hardware-aware quantization.
A second major axis of his activity connects ML theory with neuroscience and neuromorphic/in-memory computing. His recent work studies how dendritic nonlinearities and biologically grounded constraints shape capacity and learning dynamics, and he is expanding these ideas toward theory-driven training methods robust to analog hardware non-idealities (noise, drift, limited precision) and continual learning in neuromorphic hardware. In addition to academic research, he contributes to interdisciplinary and industrial initiatives at Politecnico di Milano, including collaborations in wearable/edge intelligence. He actively supervises and co-supervises PhD and MSc students and teaches in areas spanning computer science fundamentals and data-driven modeling.