Efficient modeling of uncertainty introduced by the manufacturing process is critical in the design of turbine engine components. In this study, a stochastic multiscale modeling framework is developed to efficiently account for the geometric uncertainty associated with the manufacturing process to accurately predict the performance of engine components. Multiple efficient statistic tools are integrated into the proposed framework. Specifically, a semivariogram analysis procedure is proposed to quantify spatial variability of the uncertain geometric parameters based on a set of manufactured specimens. Karhunen–Loeve expansion is utilized to create a set of correlated random variables from the uncertainty data obtained by variogram analysis. A detailed finite element model of the component is created that accounts for the uncertainties quantified by these correlated random variables. A stochastic upscaling method is then developed to form a simplified model that can represent this detailed model with high accuracy under uncertainties. Specifically, a parametric model generation process is developed to represent the detailed model using Bezier curves and the uncertainties are upscaled to the parameters of this parametric representation. The results of the simulations are then validated with real experimental results. The application results show that the proposed framework effectively captures the geometric uncertainties introduced by manufacturing while providing accurate predictions under uncertainties.