Research
My research is in the interception of DL and Computational Physics: Diffusion/Flow Matching for molecular dynamics, statistical physics for DL, and vice versa! I also think about how to ground generative models for physics.
A novel Langevin dynamics method for sampling molecular trajectories and rare events.
NeurIPS ML4PS 2024
A Diffusion-based framework for generating molecular dynamics trajectories and rare event sampling.
NeurIPS AI for Science 2023
Method for manipulating images to control activity in specific brain regions via diffusion models.
Under Review
Projects
I experiment with deep learning for art and other more creative and unorthodox applications in my spare time, including studying self-organizing systems like neural cellular automata.
An evolutionary algorithm for optimizing text-to-image prompts using aesthetic scoring.
NeurIPS ML4CD 2023
Enhanced NCA framework with multiple neighborhoods for improved texture synthesis.
ArXiv 2023
Developed a novel method using NCAs to transform images iteratively for style transfer and text-guided image transformations using a CLIP-based loss function. The method application yields good results and is applicable in real-time applications and due to the nature of NCAs is robust to overfitting.