Sara Magliacane

Sara Magliacane is an Assistant Professor at the University of Amsterdam and a Research Scientist at the MIT-IBM Watson AI Lab. She leads research at the intersection of causality and machine learning, focusing on improving AI robustness, generalization, and safety through causal reasoning. Her work spans causal representation learning, causal discovery, and causality-inspired ML in domains like reinforcement learning and dynamical systems.

• PhD in Artificial Intelligence (2017), VU Amsterdam
• MSc in Computer Engineering (2011), Politecnico di Milano/Torino
• BSc in Computer Engineering (2008), Università degli Studi di Trieste

Her research explores causal variable identification from high-dimensional data (e.g., images, sequences) and causal graph discovery for domain adaptation. Methods include CITRIS, FANS-RL, and SNAP, with applications in embodied AI and biomedical data. The group emphasizes theoretical guarantees and scalable algorithms for real-world systems.

Recent work trends include temporal causal modeling, intervention-efficient learning, and nonstationary reinforcement learning. Publications cover topics like causal discovery in partially observed settings, causal graph pruning, and sample-efficient concept learning, often combining neurosymbolic approaches with deep learning.

• Scientific Awards: ELLIS Scholar

Sara supervises PhD students across universities (UvA, University of Pisa) and collaborates with institutions like TU Delft, Harvard, and IBM Research. She co-organizes workshops at premier conferences (NeurIPS, ICML, AISTATS) and teaches causality courses at the University of Amsterdam and Harvard Data Science Initiative.

Her lab, Amsterdam Machine Learning Lab (AMLab), investigates causal structure in embodied agents, safe reinforcement learning, and hybrid dynamical system modeling. The group maintains active partnerships with institutions such as MIT-IBM Watson AI Lab, Qualcomm, and Adyen.