Stephen Green

UKRI Future Leaders Fellow · University of Nottingham

I co-lead the development of the Dingo machine learning framework for gravitational-wave inference, supported by a £1.5M UKRI Future Leaders Fellowship (2024–2028). We develop simulation-based inference methods for gravitational-wave data analysis, from parameter estimation to population studies. I also work on black hole perturbation theory, motivated by waveform modelling and foundational questions in general relativity. My group at Nottingham includes three postdocs and three PhD students, and is part of the Nottingham Centre of Gravity.

Previously I was at the Max Planck Institute for Gravitational Physics (AEI), the Perimeter Institute, and a CITA National Fellow at Guelph. I did my PhD at the University of Chicago with Robert Wald.

2026

End-to-end population inference from gravitational-wave strain using transformers

We extend DINGO from single events to population analysis: a transformer combines event-level embeddings into a single summary, and a normalizing flow decodes the population hyperparameters—about one second per catalog, no per-event PE step. Work led by Konstantin Leyde (Flatiron CCA).

arXiv Blog post

Nature · 2025

Real-time inference for binary neutron star mergers using machine learning

In work led by Max Dax (now PI at Ellis Tübingen), we demonstrate that neural posterior estimation can enable one-second inference for binary neutron star mergers—even before the merger occurs. Fast sky localization is crucial for searching for multimessenger counterparts.

Read article arXiv Nature News & Views

Software

Dingo – Deep Inference for Gravitational-Wave Observations

Open-source simulation-based inference code for gravitational waves, developed for research and use by the LIGO-Virgo-KAGRA collaboration.

Code Trained Networks