Despite continuing advances in computational power, full-body models of the human cardiovascular system remain a costly task. Two principal reasons for this cost are the total overall length of the vascular network (spanning O(10^8) m) and the broad range of length scales (from 10^−2 to 10^−6 m) involved. Multiscale modeling can be employed to overcome these issues; specifically, subsystems of higher spatial dimension representing domains of interest can be coupled at their boundaries to lower-dimensional subsystems that mimic relevant inflow/outflow conditions. Though this scheme can increase computational efficiency, the inherent reduction in spatial dimension results in parameterizations that can be difficult to optimize in patient-specific contexts. This work is divided into two parts: in the first segment, a closed-loop multiscale model of the entire cardiovascular system is developed and integrated with a feedback control model for blood pressure regulation. It is tested against clinical data for cohorts of healthy subjects, and its predictive utility is demonstrated in a simulation of acute hemorrhage from the upper leg. After validating the multiscale/reduced-order approach, a parameter optimization technique based on the ensemble Kalman filter (EnKF) is constructed. By assimilating patients’ clinical measurements, this method is shown to successfully tune parameters in two models: a zero-dimensional model of the pulmonary circulation, and a multiscale 0D-1D model of the lower leg.