Nesar Ramachandra

Nesar Ramachandra

Computational Scientist, Argonne National Laboratory

Research

My research focuses on developing and applying computational methods at the intersection of astrophysics, cosmology, and machine learning. The work spans foundation models for scientific applications, advanced ML techniques for astronomical data analysis, cosmic structure investigation, and statistical inference methods.

Disclaimer: This section is automatically updated by Reasoning Language Models. Google Gemini is utilized to periodically go over my recent publications, talks and activities to update the content. While the information is monitored, at times incorrect information may appear.

Foundation Models

Recent research highlights the transformative impact of Foundation Models on scientific inquiry, especially in astronomy. Key methods involve developing domain-specialized Large Language Models (LLMs) and multi-modal Foundation Models, often fine-tuned for specific scientific language like spectroscopy. These models apply techniques such as literature mining for concept-object association prediction and process diverse data, including cosmological simulations and ensemble data. The impact is significant: achieving expert-level performance in Q&A and scientific reasoning, often matching larger general models efficiently. These serve as intelligent research assistants, accelerating discovery, enhancing analysis, and providing interactive interpretation, underpinned by robust AI evaluation methodologies.

Learn more about this research →
Figure from Multi-modal Foundation Model for Cosmological Simulation Data
From: Multi-modal Foundation Model for Cosmological Simulation Data
8 Publications 1 Figures Available

Machine Learning for Science

Machine learning is revolutionizing scientific discovery by addressing complex data challenges across diverse domains. Key methods include deep learning, generative adversarial networks, and probabilistic modeling, often applied to tackle sparse, high-dimensional, or noisy datasets. In astronomy and cosmology, these techniques enable precise redshift estimation, advanced galaxy morphology classification, and robust detection/modeling of gravitational lenses. Anomaly detection via GANs aids in discovering novel phenomena, while neural networks enhance image deconvolution and peculiar velocity estimation. Furthermore, ML enhances interpretability through disentangled latent spaces and provides robust solutions for engineering applications like stress field analysis and global field reconstruction, fundamentally accelerating scientific understanding and discovery.

Learn more about this research →
Figure from Application of probabilistic modeling and automated machine learning framework for high-dimensional stress field
From: Application of probabilistic modeling and automated machine learning framework for high-dimensional stress field
Figure from A Modular Deep Learning Pipeline for Galaxy-Scale Strong Gravitational Lens Detection and Modeling
From: A Modular Deep Learning Pipeline for Galaxy-Scale Strong Gravitational Lens Detection and Modeling
Figure from Enhancing Interpretability in Generative Modeling: Statistically Disentangled Latent Spaces Guided by Generative Factors in Scientific Datasets
From: Enhancing Interpretability in Generative Modeling: Statistically Disentangled Latent Spaces Guided by Generative Factors in Scientific Datasets
13 Publications 9 Figures Available

Dark Matter & Cosmology

Recent Dark Matter and Cosmology research integrates diverse methods to probe the universe's structure. Cosmic shear analyses (Hyper Suprime-Cam) constrain modified gravity. Multi-stream phase-space views unveil the intricate dynamics and topological features of dark matter haloes and the Cosmic Web. Advanced cosmological simulations, like CRK-HACC, model galaxy formation, while AI/ML generative models accelerate and benchmark predictions. Observational efforts, including stellar surveys (Gaia, Red Clump) mapping the Milky Way and large-scale surveys (SPHEREx, SPTpol) tracing the Cosmic Web, further constrain cosmological parameters. This multi-pronged approach significantly enhances our understanding of dark matter and cosmic evolution.

Learn more about this research →
Figure from Multi-stream portrait of the Cosmic web
From: Multi-stream portrait of the Cosmic web
Figure from Benchmarking AI-evolved cosmological structure formation
From: Benchmarking AI-evolved cosmological structure formation
Figure from Topology and geometry of the dark matter web: a multistream view
From: Topology and geometry of the dark matter web: a multistream view
14 Publications 7 Figures Available

Emulation & Inference

Emulation and inference research leverages advanced machine learning techniques to accelerate scientific discovery from complex simulations. Methods include probabilistic neural networks, differentiable models like SHAMNet, and Gaussian process emulation, often operating within latent spaces. These surrogates enable efficient exploration of high-dimensional parameter spaces. Applications span cosmology, for subgrid models, large-scale structure, and modified gravity, to fluid dynamics, for reduced-order modeling and data recovery, and high-energy physics. The core impact lies in robust uncertainty quantification, enhanced interpretability, and significantly speeding up computationally expensive simulations, thereby enabling more comprehensive scientific inference.

Learn more about this research →
Figure from Matter Power Spectrum Emulator for f(R) Modified Gravity Cosmologies
From: Matter Power Spectrum Emulator for f(R) Modified Gravity Cosmologies
Figure from Emulator-Based Inference of Cosmological Subgrid Models
From: Emulator-Based Inference of Cosmological Subgrid Models
Figure from Probabilistic neural networks for fluid flow surrogate modeling and data recovery
From: Probabilistic neural networks for fluid flow surrogate modeling and data recovery
7 Publications 6 Figures Available