I’m interested in how LLMs learn, and how we can use that understanding to make them more efficient—through better optimizers, efficient architectures, and a bit of theory. Earlier, I worked on Bayesian inference and generative models (MCMC, variational inference, and their combinations), and later explored sequential decision-making by connecting causality and Bayesian methods. Outside of research, I love the gym 🏋️‍♀️, sports ⛹️, and relaxing with music 🎧.

Research areas: Learning dynamics, optimization, foundation models, causality, approximate inference.