Subarcsecond imaging of gas and dust in local galaxies allows us to study the formation of massive clusters, potential proto-globular clusters, on cluster scales of 10 pc or less. Massive clusters can contain thousands of O stars within a small volume and are important potential sources of feedback and enrichment in galaxies. Molecular gas masses near large clusters are difficult to constrain, even in the absence of CO-dark H2, but dynamical constraints indicate that star formation efficiencies can exceed 50%, at least an order of magnitude higher than typical Galactic values, values that are favorable to long term cluster survival. High radiation fields from young clusters with many O stars can produce warm, PDR conditions in the gas but may not disrupt the molecular cloud, which may explain how multiple generations of stars can form within a single cluster. Massive clusters create localized metal enrichment during early stages of cluster evolution, which can potentially have significant dynamical consequences for host molecular clouds and their evolution. The talk will focus on Keck and ALMA observations of massive clusters forming in the low metallicity dwarf galaxies NGC 5253 and II Zw 40 as examples of this mode of star formation.