In machine learning, I focus on problems in the domains of natural language processing (NLP) and generative modeling. Within NLP, I am most interested in large language models, especially efficient training and sparse-attention mechanisms. For generative models, I have focused on variational auto-encoders and latent-space disentanglement. A final area of focus is in advanced methods for federated learning (FL), including personalized FL, training with heterogeneous distributed data, and privacy preserving techniques.

My scientific computing research interests lie in the areas of computational mathematics and physics, specifically in the areas of finite-element methods and least squares approaches. Currently, my research focuses on the development of theoretically-supported nonlinear constrained optimization methods coupled with finite elements for static and dynamic liquid crystal simulations. Such development concerns both accurate physical modeling as well as efficient numerical computation. The coupling of liquid crystal free elastic effects with applied electric fields, flexoelectric effects, and fluid dynamics yields interesting theoretical challenges for the design of effective and efficient computational techniques. In addition, efficient algorithm development branches into adaptive mesh refinement techniques and linear solvers for saddle point systems.