University of Cambridge
academicCambridge's AI footprint is a dense cluster around data-driven scientific foundation models and probabilistic ML. The flagship line is Richard Turner's weather/climate FM program: Aardvark Weather (Nature 2025) was the first single ML model to replace the entire numerical weather prediction pipeline, and Otter (ICLR 2026) pushes the skill–compute Pareto frontier — beating IFS ENS and GenCast at a fraction of the compute — fed by a transformer neural-process method line (much of it shipped under the cambridge-mlg GitHub org).
Around it: José Miguel Hernández-Lobato's diffusion/energy-based Boltzmann samplers for equilibrium molecular distributions and diffusion methods for PDEs; Pietro Liò's geometric and topological deep learning with comp-bio FMs like gRNAde (ICLR 2025 Spotlight); and Amanda Prorok's multi-robot foundation models (CoViS-Net) plus adopted MARL infrastructure (BenchMARL, joint with Meta FAIR).
On the language side, the Language Technology Lab (Anna Korhonen, Ivan Vulić; the cambridgeltl orgs) anchors multilingual NLP and LLM-as-judge evaluation — PairS and ZEPO are reference methods for pairwise, bias-reduced LLM judging — with SapBERT (~1.64M downloads) as the legacy biomedical anchor.
Your notes
People
- Richard E. Turner Google Scholar — Professor of Machine Learning, Engineering
- José Miguel Hernández-Lobato Google Scholar — Professor of Machine Learning
- Anna Korhonen Google Scholar — Professor of AI; Co-Director, Language Technology Lab
- Pietro Liò Google Scholar — Professor, Computer Laboratory
- Amanda Prorok Google Scholar — Professor of Collective Intelligence & Robotics