Astro Automata
Automating Scientific Discovery
Our research group at the University of Cambridge develops and applies AI techniques across the physical sciences. Our work spans from foundational ML methods to scientific applications in astrophysics, cosmology, mathematics, fluid dynamics, and physics.
Our Research
Foundation Models for Science
Along with our collaborators at Flatiron Institute, we form one of the nodes of PolymathicAI, an international collaboration building large-scale foundation models for science, pretrained on diverse numerical datasets. Just as large language models have transformed natural language processing, we aim to build large "quantitative" models that transfer concepts across scientific disciplines.
Interpretable ML for Science
At the other end of the spectrum, we drive the development of machine learning methods which yield clear and interpretable scientific insights, lifting the black box predictions of deep learning. Our work includes the popular open-source framework PySR for symbolic regression, among significant work on interpretability of deep neural networks in the context of physics.
Scientific Discovery across Physics and Astronomy
We simultaneously apply our AI techniques to targeted problems across the physical sciences. Our scientific interests span a variety of topics (often student-driven), including turbulence modeling, planetary dynamics, oceanography, cosmological inference, exoplanet characterization, among many other areas.
Our Team
Miles Cranmer
PI
Assistant Professor, DAMTP & IoA
Topics: all
Meatball
Co-PI
Catbridge
Topics: Sleeping
Shruti Mishra
Encode Fellow
DAMTP
Topics: Foundation Models, ML-Accelerated Simulation
Alessandro Favero
Physics-AI Fellow
DAMTP
Topics: Foundation Models, Mechanistic Interpretability, Theoretical Physics
Tom Hehir
PhD Student
IoA
Topics: Foundation Models, Extragalactic
Rio Fear
PhD Student
DAMTP
Topics: Mechanistic Interpretability, ML-Accelerated Simulation
Júlia Laguna Miralles
PhD Student
IoA
Co-supervised with Vasily Belokurov
Topics: Time Domain, Extragalactic
Hadi Sotoudeh
PhD Student
IoA
Co-supervised with Neil Lawrence
Topics: Foundation Models, ML-Accelerated Simulation, Extragalactic
Nidhish Sagar
PhD Student
Physics
Co-supervised with Akshay Rao, James Fergusson
Topics: Symbolic ML, Energy Materials
Adil Soubki
PhD Student
DAMTP & IoA
Topics: Symbolic ML, Theoretical Physics
Ed Stevenson
PhD Student
IoA
Co-supervised with Oliver Shorttle
Topics: Exoplanets, ML-Accelerated Simulation
Elizabeth Tan
PhD Student
DAMTP
Topics: Symbolic ML, Mechanistic Interpretability
George Vassilakis
PhD Student
IoA & DAMTP
Co-supervised with Vasily Belokurov
Topics: ML-based Imaging, Extragalactic
Rachel Zhang
PhD Student
DAMTP
Topics: Foundation Models, Mechanistic Interpretability
Julia Dima
Incoming PhD Student
DAMTP
Topics: Mechanistic Interpretability
Autumn Mapes
Incoming PhD Student
DAMTP
Co-supervised with Matthew Colbrook, Tim Gowers
Topics: AI for Mathematics
Alumni
- Helen Shao, '25 (→ Harvard)
Recent Publications
Gaussian Process Latent Factor Regression for Low-Data, High-Dimensional Output Problems
In the sciences, regression tasks often require predicting high-dimensional outputs from few training examples. Multi-output Gaussian processes exc...
Emergent Transfer of a Physics Foundation Model from Simulation to Laboratory Turbulence
Whether physics foundation models can be usefully deployed on laboratory experiments remains an open question for scientific machine learning (ML)....
CTF4Nuclear: Common Task Framework for Nuclear Fission and Fusion Models
The demand for clean energy is ever increasing, with new nuclear technologies presenting a complementary solution to renewable energies. However, d...
Our Software
High-Performance Symbolic Regression in Python and Julia
Code for "Discovering Symbolic Models from Deep Learning with Inductive Biases"