#  Research Interests 

 



### 🪐 **Research**

I am a **cosmologist** working at the intersection of **fundamental physics, data analysis, and artificial intelligence**, with the goal of extracting the full physical information encoded in the **large-scale structure (LSS)** of the Universe.  
My research seeks to test the nature of **dark energy, dark matter, and gravity** by developing novel, interpretable, and computationally efficient inference methods for cosmological surveys.

The next generation of galaxy surveys, including **DESI**, the **Vera Rubin Observatory LSST**, **Euclid**, and the **Nancy Grace Roman Telescope,** will map the 3D distribution of cosmic structure with unprecedented precision. Fully harnessing this information requires moving beyond traditional two-point statistics, such as the power spectrum, which capture only a fraction of the cosmological signal. My work focuses on designing new statistical estimators and **AI-driven inference frameworks** that can efficiently extract the non-Gaussian information generated by nonlinear structure formation.

#### Wavelet Scattering and Non-Gaussian Cosmology

I pioneered the application of the **Wavelet Scattering Transform (WST)** to cosmological data, introducing a new family of summary statistics that bridge the gap between conventional clustering estimators and convolutional neural networks.  
Using state-of-the-art simulations such as **AbacusSummit**, I led the first WST-based likelihood analysis of real galaxy data from the **BOSS CMASS DR12** sample, achieving constraints up to six times tighter than the two-point correlation function. More recently, I co-led a project applying the WST and wavelet phase harmonics to cross-correlations of **CMB lensing** and **weak lensing** maps, demonstrating significant gains over standard cross-power spectra.

#### AI-Driven Inference and Simulation-Based Cosmology

As a member of the **NSF–Simons AI Institute for the Sky (SkAI)**, I am extending these ideas to **AI-based field-level inference**, leveraging machine-learning techniques such as normalizing flows and diffusion models to emulate complex cosmological fields and accelerate parameter estimation. This work aims to unify physically interpretable and data-driven approaches, enabling optimal, simulation-calibrated inference from DESI and LSST.

#### Testing Fundamental Physics

Earlier in my career, I developed **analytical and hybrid simulation techniques** to study **modified gravity theories** and their screening mechanisms, constructing fast and accurate models of structure formation in non-standard cosmologies. These methods continue to inform my current efforts to test dark-energy models, neutrino masses, and primordial signatures through large-scale-structure data.

Through these projects, I strive to build a cohesive program that connects **physical modeling, high-performance computing, and AI-based inference** to uncover the fundamental laws governing cosmic evolution.