In many areas, available seismic acquisition data may not be optimal, and purely data-driven velocity modeling methods are inadequate to resolve specific imaging problems. In this case study to image beneath a fault shadow offshore in the Gulf of Mexico, limitations of the available data were the original motivation to incorporate various interpretation-derived constraints in conjunction with data-driven tomographic velocity updates to improve the overall velocity modeling process. Automated tomographic modeling is increasingly relied upon to produce high-quality models in cases of robust data input. However, in the case of sub-optimal acquisition that limits illumination, and noise and multiples that affect the ability to pick residual moveout, interpretation constraints can enhance data-driven modeling. Strategies incorporated here include "manually seeded" velocity analysis, explicit interpretation-guided pick weighting, and implicit geologic steering filters based on dip fields. The goal of the study is to highlight effective strategies for interpretationconstrained modeling to complement data-driven modeling in an effective holistic workflow.