Scientific discovery is increasingly enabled by heterogeneoushardware that includes multiple processor types, deep memoryhierarchies, and heterogeneous memories. To effectively utilize thishardware, computational scientists must compose their applicationsusing a combination of programming models, middleware, and runtimesystems. Since these systems are often designed in isolation from each other, their concurrent execution often results inresource contention and interference, which limits applicationperformance and scalability. This problem adds to thealready complex interactions between multiple physics libraries andemerging machine learning components in scientificapplications. Consequently, real-world applications face numerouschallenges on heterogeneous machines.This poster presents Quo Vadis, an interface and runtime systemthat helps hybrid applications make efficient use of heterogeneoushardware, ease programmability in the presence of multiple programmingabstractions, and enable portability across systems. The runtimesystem abstracts out low-level details of the hardware and presents anarchitecture-independent interface applications can use to leveragelocal resources automatically and without user intervention. Theposter also includes a skeleton multi-physics application where weapplied Quo Vadis to demonstrate how the challenges described abovecan be met in a portable way across systems and with a small effortfrom application writers.