Introduction

I am currently a Junior Kavli Fellow at the Kavli Institute for Cosmology at Cambridge and a Post-Doctoral Research Associate at St Edmunds College. I am pursuing research in novel Bayesian analysis tools and their broad cosmological applications.

During my PhD I worked on data analysis techniques for global or sky-averaged 21-cm cosmology as part of the Cm-Wave Experimental Radio Cosmology group and as part of the REACH (Radio Experiment for the Analysis of Cosmic Hydrogen) collaboration under the supervision of Dr Eloy de Lera Acedo, Dr Will Handley and Dr Anastasia Fialkov.

You can find my thesis here.

21-cm cosmology is a field of study which aims to track the redshifted 21-cm emission from neutral hydrogen throughout cosmic history in an effort to determine when the first galaxies formed and what those galaxies were like. We know very little about the first luminous objects to form between the CMB and the current era of the galaxies dominated by large scale structure. While probes like JWST can in principle probe distant galaxies, as far back as z=20, 21-cm cosmology can provide statistical information about the population of galaxies over the redshift range z = 5 - 150.

While a promising probe the view to the first galaxies is fogged by emission from our own and other galaxies, emission from our own atmosphere and made challenging by the non-uniform responses of our instruments. I have worked on a number of these challenges during my PhD including; galactic and extragalactic foreground modelling using maximally smooth functions (see maxsmooth), signal emulation (see globalemu) improving on the runtime of the state of the art emulators by a factor of 102 and the application of Normalizing Flows in a machine-learning enhanced Bayesian workflow (see margarine).

You can read about the application of the tools I have developed to data from SARAS2 here, data from SARAS3 in Nature Astronomy and data from HERA and SARAS3 here.

I have recently worked on a novel approach for learning multi-modal probability distributions with Normalizing Flows which you can read about here.