Towards Energy-Efficient Split Computing: A Hardware-Software Co-Design Perspective

This extended abstract was accepted for presentation at the poster session of EuroSys 2025, the European Conference on Computer Systems, held from March 30 to April 3, 2025, in Rotterdam, Netherlands. It outlines a framework for deploying machine learning models on edge devices using split computing and hardware-software co-design. Our approach dynamically selects optimal computation splits and hardware configurations to minimize energy consumption while meeting latency constraints for inference. The poster accompanying this abstract is available here.