The goal of my research is to enable informed design-phase decisions that improve the energy efficiency and performance of machine learning models across different hardware platforms. By analyzing potential bottlenecks early on, applications can be made more resource-efficient from the start.