My research aims to create more practical machine learning systems. I particularly focus on the inputs and outputs, which are underexplored. On the input side, this means learning paradigms that can better leverage the full breadth of available data. On the output side, this means improving generalization and shrinking the gap between laboratory measurements and real world outcomes. The research I’ve done on these themes bridges multiple domains and applications, including work towards reliable generative AI, healthier social media ecosystems, less human trafficking, and more thoroughly evaluated machine learning systems.