Recent years have witnessed a dramatic increase in computational fluid dynamic (CFD) simulations for diagnosing cardiovascular diseases, which continue to dominate healthcare costs and are projected to be over one trillion dollars by 2035. Current frameworks, however, face three key technical challenges: simulations are memory intensive with high time-to-solutions; the need for validation against in vivo measurements; and methods for clinicians to intuitively interact with the simulation results are lacking. In this thesis, we overcome these challenges by first establishing a novel, memory-light algorithmic representation that both reduces the memory requirements by 74% and maintains excellent parallel scalability. Second, we validate our CFD framework through a multicenter, clinical study comparing invasive pressure measurements to calculated values for 200 patients. Third, we assess how physicians interact with large-scale CFD simulation data and present a virtual reality platform to enhance treatment planning. We expect this work to lay the critical groundwork for translating the use of massively parallel simulation-driven diagnostics and treatment planning to the clinic. Our long-term goal is to enable the use of personalized simulations to improve clinical diagnosis and outcome for patients suffering from cardiovascular diseases.