Accurate patient-specific modeling of cardiovascular flows is crucial for reliable predictions of hemodynamic parameters, such as wall shear stress, which play a central role in assessing cardiovascular diseases like atherosclerosis. However, in-vivo velocity boundary data, often derived from modalities like 4D Flow MRI, are frequently limited by low resolution and noise, complicating precise simulations. To overcome these challenges, we propose a novel framework that combines computational fluid dynamics with a stochastic data assimilation technique[1,2]. This method enhances boundary estimation accuracy by iteratively incorporating velocity data into the vascular model over time, enabling real-time refinement of unknown boundaries. Our simulations, governed by the incompressible Navier–Stokes equations, evaluate constant, time-dependent, and time-space-dependent boundaries in both two- and three-dimensional models. Results demonstrate significant error reductions, with relative errors as low as 0.996% for constant boundaries in 2D and 7.37% for space-time-dependent boundaries in 3D patient-specific models. By improving the accuracy of velocity boundary profiles, our approach enhances wall shear stress predictions and contributes to more reliable cardiovascular flow modeling. These advancements hold promise for better diagnostic and therapeutic strategies in patient-specific cardiovascular care.