A generative deep learning model of molecular structure is combined with supervised deep learning models of molecular properties to achieve high-throughput (multi-)property-driven design of organic molecules.