Computational models are increasingly developed to facilitate clinical decisions. In order to integrate a computational model into a clinical workflow, it needs to demonstrate reliable predictions under uncertainty, and patient inter- and intra-variability. We present a modeling pipeline encompassing model generation, calibration to patient-specific data, estimation of parameter ranges, sensitivity analysis, inverse problem solving, identifiability analysis, and uncertainty quantification. The pipeline begins with the formulation of a physiologically interpretable reduced-order model, followed by calibration to available measurements with the CMA-ES algorithm, an optimization-based inverse method. Parameter ranges are defined based on prior knowledge and population variability, forming the basis for global sensitivity analysis to identify influential parameters. Computationally intensive sensitivity analysis and uncertainty quantification can be accelerated thanks to surrogate models, enabling efficient exploration of high-dimensional parameter spaces [1]. Inverse problem analysis and uncertainty propagation are then the basis to quantify confidence in model predictions, while structural and practical identifiability analyses determine which parameters can be reliably inferred from the available data. This integrated framework supports transparent model reduction, robust parameter inference, and credible predictive simulations, providing a systematic approach to patient-specific modeling under uncertainty. As an application, the framework is demonstrated on a closed-loop lumped-parameter model of whole-body hemodynamics in patients with liver cirrhosis undergoing hepatic interventions such as transjugular intrahepatic portosystemic shunt placement. The circulation model represents with Windkessel elements the pulmonary circulation, the digestive organs, the liver and the remaining systemic circulation. Cardiac function is modeled by the single-fiber heart model, which provides physiologically interpretable parameters and is easier to calibrate [2]. Simulations are performed within LumpedFlux, a novel in-house framework for 0D hemodynamic modeling. Patient-specific calibration enables prediction of post-interventional hemodynamic responses, illustrating the framework’s potential for intervention planning and risk stratification in liver disease treatment. [1] Hanna et al., Comput. Biol. Med 2025 [2] Haghebaert et al., J. Physiol. 2025