Because great earthquakes in the Himalaya have an average recurrence interval exceeding 500 yr, most of what we know about past earthquakes comes from paleoseismology and tectonic geomorphology studies of the youngest fault system there, the Main Frontal Thrust (MFT). However, these data are sparse relative to fault segmentation and length, and interpretations are often hard to validate in the absence of information about fault geometry. Here, we image the upper two km of strata in the vicinity of the fault tip of the MFT in central Nepal (around the town of Bardibas) applying a pre-stack migration approach to two new seismic reflection profiles that we interpret using quantitative fault-bend folding theory. Our results provide direct evidence that a shallow décollement produces both emergent (Patu thrust) and blind (Bardibas thrust) fault strands. We show that the décollement lies about 2 km below the land surface near the fault tip, and steps down to a regional 5 km deep décollement level to the north. This implies that there is significant variation in the depth of the décollement. We demonstrate that some active faults do not reach the surface, and therefore paleoseismic trenching alone cannot characterize the earthquake history at these locations. Although blind, these faults have associated growth strata that allow us to infer their most recent displacement history. We present the first direct evidence of fault dip on two fault strands of the MFT at depth that can allow terrace uplift measurements to be more accurately converted to fault slip. We identify a beveled erosional surface buried beneath Quaternary sediments, indicating that strath surface formation is modulated by both climate-related base level changes and tectonics. Together, these results indicate that subsurface imaging, in conjunction with traditional paleoseismological tools, can best characterize the history of fault slip in the Himalaya and other similar thrust fault systems.