Power Constrained Bandits
How do we guarantee sufficient statistical power under real-life constraints?
Lots of people find my first name hard to pronounce. Usually, I don't mind whichever you say it. In case you are really interested,
here is the correct pronunciation ().
I am currently a postdoc advised by Prof. Barbara Engelhardt at Gladstone Institutes. Before my postdoc, I obtaind my PhD at the Data to Actionable Knowledge (DtAK) Lab at Harvard University, advised by Prof. Finale Doshi-Velez. My research focuses on identifying and tackling challenges along the machine learning pipeline -- from data preparation to model deployment -- with the goal of bridging the gap between machine learning methodology and clinical application. Specifically, my current research includes developing pipelines for data processing and exploration, developing models that are robust under real-life limitations (e.g. data heterogeneity, data scarcity and domain knowledge constraints), as well as developing quantitative and qualitative assessments of models for specific downstream desiderata (e.g. statistical power, human-interpretability). Prior to joining DtAK, I received my bachelor’s degree in applied math and computer science from Emory University, where I was advised with Prof. Joyce C Ho.