Cognitional heart disease is one of the leading case of mortality among newborns. Tissue-engineered vascular grafts presents the potential to help treat newborn with cognitional heart disease. This paper presents a computational framework for the automatic shape optimization of patient-specific tissue-engineered vascular grafts for repairing the aortic arch, aiming to improve the current treatment outcomes. The core innovation lies in an automatic shape optimization pipeline that combines Bayesian optimization techniques with the open-source finite volume solver, OpenFOAM, and a novel graft deformation algorithm. In this study we used imaging and flow data obtained from six patients diagnosed with cognitional heart disease. Our framework begins with Laplacian mode computation and the approximation of a computationally low-cost Gaussian process surrogate model to capture the minimum of weighted combination of inlet-outlet pressure drop (PD) and maximum wall shear stress (WSS). Subsequently, with a limited number of OpenFOAM simulations, the optimal shape is identified. The result showcases the potential of online training and hemodynamic surrogate model optimization for providing optimal graft shapes. The resulting graft designs for six patient-specific cases minimize PD and WSS, offering improved hemodynamic performance compared to initial models. These results show our framework successfully reduce inlet-outlet PD and maximum WSS as key factors influencing hemodynamic performance in comparison to baseline models. Furthermore, we compare how the performance of each design optimized under steady-state simulation compares to its performance when simulated under transient simulations. Our findings underscore that the automated designs achieve at least a 17% reduction in blood flow pressure drop in comparison to the baseline model, establishing their superiority over geometries optimized by human experts. Also, we evaluate to what degree optimal design shapes computed based on steady-state simulations remain performant under transient simulations.