Research

My research lies at the intersection of cosmology and astrophysics, driven by a broad interest in understanding the large-scale structure of the Universe and its evolution across cosmic time. I focus on developing and applying both physical and data-driven tools to extract information from high-redshift observables such as the Lyman-α forest and Lyman-α emitters (LAEs), with the overarching goal of constraining the physics of the intergalactic medium (IGM), cosmic reionization, and the growth of structure.

A major area of focus has been the Lyman-α forest, which traces the diffuse intergalactic medium (IGM) and provides a powerful probe of cosmological structure and astrophysics on small and intermediate scales. I led the first observational measurements of the three-point correlation function of Lyman-α absorbers, both at low redshift (z < 0.5) using HST-COS data and at high redshift (1.7 < z < 3.5) using Keck-HIRES and VLT-UVES spectra. These measurements allowed for a first look at the evolution of non-Gaussianity in the distribution of intergalactic gas across cosmic time, providing new insights into the anisotropies and velocity structure of the IGM.

To advance inference from such data, I began applying machine learning techniques, particularly in the form of Information Maximizing Neural Networks (IMNNs), which are designed to compress data into informative summaries for parameter estimation without relying on explicit likelihoods. I demonstrated that these approaches can extract robust astrophysical and cosmological parameters from the Lyman-α forest in comparison to traditional approaches.

I have also studied the clustering of metal absorbers such as O VI, using simulations to understand how feedback processes and fluctuations in the UV background impact their spatial distribution. These systems offer a complementary view of the IGM and its enrichment history, bridging the gap between the Lyman-α forest and galaxy environments.

More recently, I introduced the LAE bispectrum as a novel probe of the morphology of reionization, showing that it is sensitive to the topology of ionized regions and the underlying source models. This higher-order statistic provides complementary information to the LAE power spectrum and enhances the utility of LAEs as tracers of the Epoch of Reionization.

I also developed DeepCHART, a deep learning framework for tomographic reconstruction of the 3D dark matter field from sparse observational tracers such as Lyman-α forest skewers and galaxy positions. DeepCHART is designed to be generalizable across observables, scalable to large volumes, and robust to sparse or noisy input. While not tied to any specific redshift regime, it enables high-fidelity recovery of the cosmic web and is well-suited for application to both current and upcoming surveys.

Together, these projects form a cohesive effort to build interpretable, data-efficient models of the high-redshift Universe, by combining rigorous physical insight with modern machine learning and simulation tools.