Carnegie Postdoctoral Fellow
Computational Data Analysis
I work on applying computational data analysis techniques to interesting astronomy problems. Currently I'm working on building a 3D dust map of the Milky Way galaxy using scalable Gaussian processes. I am also interested in falsifying dark matter models using stellar streams around the Milky Way galaxy observed by Gaia.
I enjoy thinking about data-driven approaches in situations with complex physical models, where I want to build the model from the data rather than the physics. I apply these concepts to stars using the Gaia catalogue where our physical understanding of stars and stellar models aren't sufficient for the precision we need and introduce too much bias. I made the Gaia DR1 parallaxes more precise using a data-driven model of stars.
I enjoy combining data-driven approaches with physical models where we want to interpret the results, like measuring the shape of the dark matter halo using stellar streams. Stellar streams are a powerful tool to falsify dark matter models, but these models, especially using multiple stellar streams in a time dependent potential, require many parameters. I am currently working on implementing this as a tractable model.
I also enjoy thinking about making complex models computationally tractable which requires advances in computational efficiency. Each prediction of a complex model is expensive to generate, and for inferences we need a continuous prediction over a wide range parameter space. I took a step towards this goal by implementing a method that reduces sample variance in predicted power spectra. This quicker convergence of the mean predictions frees up computing time for exploring parameter space.