The objective of this program is to develop a data-driven approach to synthesis science by combining text-mining and machine learning, in situ and ex situ characterization of experimental synthesis, and large-scale first-principles modeling. Data extraction and natural language processing on the corpus of scientific literature provides a comprehensive dataset representing the historic facts on inorganic synthesis. These are used to machine-learning synthesis suggestions from, and to test and validate microscopic models and theories for predicting synthesis pathways. Hypotheses constructed from data and fundamental science ideas are tested in detail against systematically planned and executed experiments in autonomous labs in an active learning loop. Synthesis pathway hypothesis are evaluated mechanistically through in situ and ex situ characterization of the synthesis process. Through these efforts, one expects to obtain a quantitative understanding of how several key factors related to thermodynamics and kinetics influence synthesis outcomes.