Raia Hadsell
VP of Research and co-lead of Frontier AI at Google DeepMind
A deep-learning researcher whose work spans representation learning, robot navigation, and methods that help neural networks retain previously learned skills.
Biography
Raia Hadsell is a vice president of research at Google DeepMind and co-leads its Frontier AI unit. Her path into computing began after undergraduate study in religion and philosophy at Reed College. She completed a doctorate at New York University with Yann LeCun, working on neural representations and vision for mobile robots.
Hadsell co-authored research on learning invariant mappings and later investigated navigation, transfer learning, and continual learning. Her work at DeepMind included progressive neural networks and elastic weight consolidation, approaches addressing how a network can learn new tasks without losing useful abilities acquired earlier.
She joined DeepMind in 2014 after research roles at Carnegie Mellon and SRI International. Beyond research, she helped establish Transactions on Machine Learning Research and has served in organizations supporting the machine-learning community. Her contribution combines technical methods with research leadership, particularly around learning systems that must adapt over time rather than solve one fixed task in isolation.
Notable contributions
Co-authored invariant-mapping research
Developed an approach for learning representations that preserve useful similarity structure.
Contributed to continual-learning methods
Co-authored progressive-network and elastic-weight-consolidation work addressing transfer and catastrophic forgetting.
