We also develop generative AI (GenAI) methods that solve a so-called inverse design problem. These methods enable the design of molecules with the desired physicochemical, biological, or both properties. We successfully validated this approach by designing small-molecule protein kinase, protease, and GPCR inhibitors. We showed that a trained AI model can mimic expert chemists’ skills. This is a prime example of the transfer of decision power from human experts to AI. Using CMU Cloud Lab, we are developing a novel ML-guided molecular discovery platform that combines synergistic innovations in automated experiments and automated machine learning (AutoML) and Reinforcement learning (RL) agents. A software-controlled synthesis platform enables rapid iterative experimental–computational cycles and explores non-intuitive design criteria identified by an agent.