AI/ML Evaluation Frameworks

How can we make sure a machine learning algorithm is doing what we want it to do, at the performance of what we expect? We develop new methodologies for rigorous and safe evaluation of machine learning under both randomized and observational settings with minimal assumptions.