Additive manufacturing of materials always involves non-equilibrium processes that lead to complex co-evolution of the thermal field, microstructure, and most importantly, residual stresses, which give rise to internal defects and significant part distortion. However, a robust strategy for controlling the level and distribution of residual stresses remains elusive. Here, we present a theoretical and experimental framework for residual stress control, in which kinetics, mechanics, and data-driven models are integrated, and are validated by in situ thermal imaging and ex situ dimension measurement. 3D finite element modeling of the thermo-mechanical process captures the underlying physics; Kriging meta-modeling of the process-property relationship accelerates computation speed; Bayesian calibration identifies model discrepancies and enables predication with different geometric designs and processing parameters. This framework enables residual stress control on the component scale with low computation cost, which is demonstrated in processes such as fused deposition modeling and directed energy deposition.