Juan Cervino is an MIT-IBM Distinguished Postdoctoral Fellow in electrical engineering whose research is focused on developing intelligent systems that are efficient under scale, robust to perturbations, and flexible to heterogeneity. Specifically, Juan’s work concentrates on the introduction of requirements to perform machine learning over graphs. As a doctoral student, Juan’s notable achievements included the development of a methodology to train graph neural networks on a sequence of growing graphs that converge to the limit architecture (the graphon) and a novel approach to learning smooth functions over the underlying low-dimensional representation of the data (the manifold). With the support of a postdoctoral fellowship, Juan will pursue several areas of inquiry, including new approaches to large-scale continual learning problems, utilizing manifold smoothness to optimize learning-for-control, and finally, expanding our understanding of collaborative solutions under heterogeneity for the purposes of developing methods of jointly learning shared solutions that improve all the agents in a network. Juan’s innovative research has strong potential to advance the field of machine learning and help deliver next-generation intelligent systems with greatly improved efficiency, strength, and flexibility.