Fellow Detail

Applied deep learning, astronomy, software development (naturally)

I'm a postdoc at the Jodrell Bank Center for Astrophysics, Manchester, where I use deep learning research breakthroughs to solve astrophysical questions. I am currently working on creating algorithms to interpret radio images of the gas streaming out of supermassive black holes, and on understanding the origin of ""fast radio bursts"" - mysterious millisecond pulses of radio energy coming from outside our galaxy. I am also the lead data scientist for Galaxy Zoo (www.galaxyzoo.org), a citizen science project using hundreds of thousands of volunteers to measure the shapes of millions of galaxies.

I did my PhD at Oxford, working on combining citizen science and machine learning to do better science than either alone. Much of my time focused on combining Bayesian deep learning and active learning to make efficient use of volunteer's time, in part as a machine learning backend for Galaxy Zoo.

I'm interested in the gap between how machine learning algorithms behave in idealized computer science contexts and in their practical application. This gap comes from several fundamental concerns - robustness, interpretability, calibration, bias, etc. - and addressing these is central to my work.