Towards Sustainable Machine Learning with Serverless at the Edge

Energy consumption has received significant attention in data centers and edge environments. The growing demands of AI/ML workloads, which are highly computational and energy-intensive, further increase the importance of energyaware design. Although energy consumption has been widely studied from hardware and software perspectives to optimize it in cloud and edge environments, the rise of serverless computing presents new challenges for applying its metrics within adaptive and sustainable systems. Our research analyzes function-level energy consumption to support an adaptive, energy-aware system. This poster compares energy usage across different workloads, highlighting how application characteristics influence total energy consumption. Building on these insights, we propose a framework that enables intelligent function selection based on energy and performance trade-offs.