Broadly speaking, my research has three (significantly) overlapping themes:

  1. Machine Learning on non Euclidean domains, Graph Signal Processing, Riemannian Geometry:
    Exploiting non Euclidean structure in datasets for better inference, data exploration, and learning. I have also been exploring the use of graph signal processing in various ML applications.
  2. Computer Vision
    I am actively pursuing research in the following : traditional inverse problems, exploiting generative adversarial networks for various learning tasks, interpretability of deep neural networks, building “better” latent spaces.
  3. Machine Learning for X
    I collaborate with scientists from a variety of disciplines providing machine learning solutions to their domain specific problems. I am currently working with data from: computational biology, geophysics, high energy physics, high performance computing, and non destructive imaging.