Text this: Deep reinforcement learning enables conceptual design of processes for separating azeotropic mixtures without prior knowledge.