Large amounts of scientific data remain uncurated, especially small datasets, which are currently invisible or "dark data". This dark data should be curated locally with the involvement of non-scientists at long-lived institutions like libraries and museums that have experience managing scholarly information over time. New roles and skills are needed for data scientists, digital curators, and biological information specialists to help address this problem by developing the necessary cyberinfrastructure, data standards, and educational programs to make more scientific dark data accessible.