Conventional Bayesian calibration often assumes the computational model can replicate the true mechanism behind data generation. However, this idealization is generally not achieved for physical models since they carry misspecification due to different parameterizations and assumptions. While accounting for data noise and parametric uncertainties has become routine in the uncertainty quantification context, the same can not be said about model errors, also known as model structural error, or model misspecification. Ignoring model error can lead to overconfident calibrations and poor predictive capability, even when high-quality data are used. I will demonstrate a stochastically embedded model correction approach that enables predictions while respecting the underlying physics. Employing polynomial chaos expansions to represent the correction terms, the approach allows the physical model to be treated as a black-box, followed by efficient quantification, propagation, and decomposition of uncertainty that includes contributions from data noise, parameter posterior uncertainty, and model error. Time permitting, the key strengths of this method will be demonstrated on realistic engineering applications including chemical kinetics, climate models and turbulence modeling.