Research

My research focuses on understanding the intricate design of the cosmic web and its implications for fundamental physics. I develop and apply a multi-stream view to unveil the caustic architecture, topology, and internal dynamics of dark matter halos and the cosmic web. This methodology provides a detailed phase-space portrait, elucidating how dark matter streams converge to form the structures we observe. Concurrently, my work also involves creating efficient emulators for the matter power spectrum in modified gravity cosmologies, such as f(R). This allows us to rapidly predict observables and test deviations from standard gravity, linking the precise structural characteristics of the universe to fundamental theories of gravity.

Learn more about this research →

My research focuses on leveraging artificial intelligence to transform cosmological simulations. I develop methods for benchmarking AI-evolved cosmological structure formation, ensuring physical fidelity and statistical accuracy against traditional N-body simulations. My work also includes building differentiable models, like SHAMNet, which provide robust predictions for large-scale structure and facilitate inference. Furthermore, I pioneer multi-modal foundation models for diverse cosmological simulation data, offering a unified framework for analysis and accelerating scientific discovery. Our rigorous physical benchmarking ensures AI-generated cosmic web features adhere to fundamental physical principles, pushing the boundaries of simulation capabilities and observational comparisons.

Learn more about this research →

My research focuses on leveraging advanced AI and deep learning to revolutionize observational astronomy. I develop methods for precise photometric analysis, mapping Galactic structure using millions of Red Clump stars. My work integrates generative adversarial networks for robust anomaly detection in vast astronomical image datasets, enabling novel cosmic discoveries. I apply machine learning to spectroscopy for probabilistic redshift estimation and train large language models to interpret complex spectral information. We utilize deep neural networks for peculiar velocity estimation from the kSZ effect and optimize galaxy selection for accurate weak lensing cluster mass measurements, advancing understanding of cosmic structure and evolution through automated, high-fidelity data analysis.

Learn more about this research →

My research focuses on advancing AI/ML methodologies for complex scientific and engineering problems. I develop probabilistic neural networks and reduced-order models to create accurate, uncertainty-aware surrogates for high-dimensional systems like fluid flows and stress fields. Our work also explores latent-space representations, employing techniques like statistically disentangled latent spaces guided by generative factors to enhance interpretability in generative modeling. Furthermore, I apply Gaussian process emulation for non-intrusive ROMs and address global field reconstruction from sparse sensor data using Voronoi tessellation-assisted deep learning. This enables robust data recovery and efficient analysis of intricate physical phenomena.

Learn more about this research →