About me

I’m a doctoral researcher at UCL’s Centre for Doctoral Training in Data Intensive Science working with the Scientific AI (SciAI) group at the Mullard Space Science Laboratory. My thesis supervisors are Prof Jason McEwen and Dr Benjamin Joachimi. My research focuses on developing machine learning and statistical methods for Astrophysics and beyond.

If you’re interested in my work or have any questions please drop me an email!

Projects

harmonic

harmonic is an open source, well tested and documented Python implementation of the learned harmonic mean estimator to compute the marginal likelihood (Bayesian evidence), required for Bayesian model selection. It is available on GitHub and PyPi. My recent project focuses on introducing normalizing flows to learn the importance sampling target distribution, increasing robustness and scalability, as described here. The original paper introducing the learned harmonic mean is available here.

S2WAV: Differentiable and accelerated spherical wavelets with JAX

S2WAV is a JAX package for computing wavelet transforms on the sphere and rotation group, available on GitHub and PyPi. It leverages autodiff to provide differentiable transforms, which are also deployable on modern hardware accelerators (e.g. GPUs and TPUs), and can be mapped across multiple accelerators. The paper introducing the algorithms is available here.

Coreference resolution for quote attribution at the Guardian

As part of my industry group project I had the opportunity to work with the amazing Data Science team at the Guardian on a project on coreference resolution for quote attribution in Guardian articles. Our work is described in a blog post on the Guardian website: “Who said what: using machine learning to correctly attribute quotes”, here.

Alan Turing Institute Data Study Group

Intense research sprint working on detecting land cover change in the Peak District National Park using multispectral aerial photography. Our team produced change maps using statistical and machine learning methods, explored the use of Siamese networks, autoencoders and introducing synthetic change to train supervised models for the task. Results from the project are described in a report available here.