As high-performance computing (HPC) systems grow increasingly heterogeneous and application workloads become more diverse, achieving both high performance and energy efficiency poses a significant challenge. While performance optimization remains essential, energy efficiency has emerged as a critical priority due to substantial infrastructure demands, operational costs, and environmental impact. This project tackles two fundamental barriers to energy-conscious HPC: power waste in heterogeneous hardware and the lack of dynamic power coordination. It develops a holistic framework, EcoHPC, to enable energy-efficient execution of hybrid workloads on heterogeneous systems. The anticipated outcomes have broad scientific, economic, and environmental impacts. Additionally, an integrated education plan aims to train the next generation of the HPC workforce. The project introduces three core technical innovations. First, it exploits collaborative filtering-based recommendation systems that combine offline analysis with real-time profiling to model application performance-power trade-offs and guide scheduling decisions. Second, it applies multi-objective optimization and Pareto-front analysis to treat power as a first-class schedulable resource, enabling system-wide coordination and optimization. Third, it develops adaptive runtime systems that dynamically predict application phases and resource demands, allowing applications to minimize power waste while maximizing performance under