MS Materials Science @ Georgia Tech (4.0 GPA) | Building generative AI for materials discovery I work at the intersection of deep learning and computational materials science — building closed-loop systems where AI generates candidate materials and simulation validates them.
- polymer_diffusion — Conditional discrete diffusion LM for property-conditioned polymer generation (ModernBERT + GaussianFourierProjection + CFG). Rivals PolyT5.
- polymer_mcts_diffusion — MCTS-guided bidirectional diffusion with confidence-driven unmasking + xTB quantum oracle. Self-play explores beyond training distribution.
- polymer_mcts_transformer — AlphaZero-style MCTS + autoregressive GPT transformer with AdaLN for polymer discovery.
- grain_growth_u-net — 5 diffusion model architectures for predicting grain microstructure evolution.
- PINN — Physics-Informed Neural Networks for meshless PDE solving. 1000x accuracy over FDM.
PyTorch Diffusion Models Transformers MCTS LAMMPS DFT (xTB) VASP HPC (A100/H100/H200) SLURM ANSYS COMSOL
🤝 Always open to collaborating with researchers, ML engineers, and deep-tech founders working on AI4Science.