Hemodynamic analysis is widely adopted in the coronary medical device space. Adoption of high-fidelity analysis of wall shear stress, turbulence, stent apposition, and fluid flow fields have high potential for guiding clinical decision making. Computational hemodynamic models currently available in the clinic remain limited by long computation times requiring substantial simplification for deployment in urgent clinical cases. Surrogate models trained using forward simulations have emerged as a viable option to provide high-fidelity simulations at orders of magnitude lower computational cost, potentially enabling deployment in clinical settings. This paper proposes and demonstrates a possible V&V approach for ML-derived surrogate models that is consistent with the philosophy inherent in the conventional sense.