Leonard Berrada

Leonard Berrada is a Research Scientist at DeepMind, focusing on robust and verified AI. He previously completed his DPhil/PhD at the University of Oxford under Andrew Zisserman and M. Pawan Kumar.

Education:

• University of Oxford (DPhil/PhD, supervised by Andrew Zisserman and M. Pawan Kumar)

Research Interests: His work spans optimization, deep learning, verification, and privacy-preserving machine learning. Recent publications highlight expertise in federated learning, biomedical image analysis, inverse problems, and tumor growth modeling.

Publication Trends: His recent work (2022–2025) focuses on federated learning, biomedical image analysis, and inverse problems in personalized tumor modeling. Earlier contributions (2020–2018) address neural network training, verification, and loss function design. Keywords include Computer Science, Optimization, and Biomedical Engineering.

Scientific Awards:

• Outstanding Reviewer Award at CVPR 2018
• Top 10% Reviewer at NeurIPS 2020

Notable Contributions: Funded by Yougov and EPSRC during his PhD, he pioneered methods in neural network verification and federated learning. His thesis was accepted without correction by Andrea Vedaldi and Julien Mairal (2020). He also contributed to Gilbert Strang's book on linear algebra and learning from data.