Research
My research combines machine learning, Bayesian statistical inference, and high-performance computing to solve hard data problems in cosmology. The unifying thread across all my projects is the challenge of extracting precise measurements from complex, noisy, high-dimensional data - often in settings where the forward model (i.e., going from theory to prediction) is expensive, approximate, or both.
The scientific context is galaxy cluster cosmology. Galaxy clusters are the largest gravitationally bound structures in the Universe, and their abundance as a function of mass and cosmic time is extremely sensitive to the fundamental parameters governing the evolution of the Universe. Building a cosmological experiment around clusters therefore requires solving several intertwined inference problems: detecting clusters in noisy telescope data, estimating their masses from indirect proxies, characterizing measurement systematics, and combining all of this within a statistically principled framework.
I am particularly interested in clusters detected via the thermal Sunyaev-Zeldovich (tSZ) effect - a spectral distortion of the cosmic microwave background (CMB) caused by hot gas in and around clusters. The tSZ signal is unique in that it does not fade with distance, making it a powerful probe for building large cluster catalogs spanning wide ranges of redshift. Exploiting these catalogs for cosmology requires careful modeling of cluster gas physics and tight control of observational biases - problems I address through the methods below.