Machine learning and data-driven science complement traditional theory, empirical and computational scientific paradigms. Machine learning and statistical approaches lie at the heart of the modern scientific research pipeline from data collection (design of experiments), to driving experiments and conducting analysis (inverse problems and optimization). Our research activities span all of these themes across applications in drug discovery, microbiome studies, particle physics, quantum physics and molecular biology. On the methods side, key areas of focus include developing interpretable machine learning models that align with scientific principles, creating robust algorithms capable of handling noisy and heterogeneous data, and integrating domain knowledge with data-driven approaches.