Core methodological work on probabilistic modeling, interpretability, and efficient training for inverse problems.