My Ph.D. research focuses on AI-driven materials science.

As a Ph.D. candidate in Computer Science at the University of South Carolina, I conduct research at the intersection of AI/ML and materials science. My work focuses on developing advanced machine learning methods to accelerate the discovery and design of novel materials. I specialize in graph neural networks (GNNs), transformers, and diffusion models for solving core materials informatics challenges such as crystal structure prediction (CSP), materials property prediction, and generative modeling of materials.

My research particularly emphasizes AI-driven approaches for crystal structure prediction — a long-standing challenge in materials science similar to the protein structure prediction problem. I draw inspiration from breakthroughs such as DeepMind's AlphaFold2 (won 2024 Nobel Prize in Chemistry) and aim to extend similar paradigms to crystalline systems. By integrating scientific domain knowledge with state-of-the-art AI models, I seek to enable automated materials discovery and accelerate innovation in materials design.

My current research focuses on developing a diffusion model for conditional generation of 3D crystal polyhedra (building blocks of crystals), given their chemical compositions (e.g., NaCl, SrTiO₃, Cu₄FeGe₂S₇). Successful prediction of the polyhedra of a crystal can lead to accurate prediction of the full crystal structure and provide an efficient solution to the CSP problem. I am inspired by successful diffusion model-based approaches in similar problems, such as molecular and protein structure prediction.