BioSara is a PhD student at CDS, co-supervised by Jonathan Niles-Weed and Rajesh Ranganath. She’s interested in the intersection of optimal transport and generative models: both flow-based and transformers. It inspires her when the improvements in deep learning are coming from a mathematical theory. Sara was previously affiliated with Gatsby Computational Neuroscience Unit at UCL, where she worked on mechanistic interpretability as a part of Saxe Lab. She finished her undergraduate studies in Mathematics and Computer Science at University of Oxford, where she focused on stochastic analysis and ML.