Abstract Molecular simulations are one of the most prominent discovery tools in science and engineering, widely adopted in applications ranging from drug discovery to materials design. The fidelity of the predictions of molecular simulations hinges on the reliability and accuracy of the chosen force fields. Thus, the quantification of uncertainties associated with the form of force fields and their parameters is a fundamental part of molecular modeling. This aspect has been rather overlooked in the past but constitutes today an active field of research. This review focuses on the data-driven Bayesian calibration of force fields. The calibration of molecular force fields is a very challenging task in the presence of an ever-increasing amount, type, and scale of reference data. Bayesian data analysis provides an invaluable tool to address this challenge, notably in the presence of model inadequacy. It enables a rigorous definition of uncertainties for the force fields and allows for data-driven robust predictions through molecular simulations. This chapter reviews recent methodological and algorithmic developments in Bayesian calibration of force fields and their application to atomic and molecular fluids. We discuss Bayesian approaches and their computational challenges followed by a description of target applications. We believe that the advances in Bayesian inference presented in this chapter would be valuable for a number of other fields in science and engineering that are revolutionized by data science.