My group's main research is studying how galaxies formed their stars. Past projects involved reconstructing star formation histories for all observable galaxies as a function of their halo mass and redshift, as well as making empirical connections between galaxies and their host dark matter halos. At present, we are refining these models to better model galaxies' colors, dust content, and metallicities.

Our group has also developed methodology for reconstructing growth histories also applies to black holes. Quasar luminosity functions combined with AGN occupation fractions and z=0 black hole mass functions provide constraints not only on black hole accretion histories but also on their typical Eddington ratios and duty cycles as a function of redshift and host galaxy mass.

Deep neural networks allow self-consistent inferences from multiple data sources that was not possible before. My group uses machine learning to measure halo properties beyond mass (e.g., accretion rates, concentrations, and spins) from combining many observable features simultaneously (e.g., satellite angular momentum distributions and galaxy environments). We are also developing new methods for Bayesian Deep Learning.

I am the main developer of the Rockstar (Robust Overdensity Calculation using K-Space Topologically Adaptive Refinement) phase-space halo finder. Rockstar excels at identifying halos and substructure where other halo finders often fail---in major mergers and at the centers of large clusters.