Angelo
Porrello.

I am a Tenure-Track Assistant Professor at the University of Modena and Reggio Emilia and a member of AImageLab.

My research challenges the idea that a single model should solve every task. Instead, I focus on model plurality: specialized models optimized for different objectives, whose capabilities can be updated, combined, and transferred.

My interests include modular deep learning—especially model merging and task arithmetic—training for model compositionality (e.g., linearized fine-tuning), and continual and federated learning, mainly in computer vision.