Software Design Decisions for Greener Machine Learning-based Systems

The widespread integration of Machine Learning (ML) in software systems has brought forth unprecedented advancements, yet the surge in energy consumption raises ecological concerns. This research addresses the environmental impact of ML development, focusing on the energy implications of design decisions in ML-based systems. This thesis aims to offer insights into the energy consumption patterns influenced by deployment architecture and training environment. Different case studies on ML-based systems will be conducted to validate and demonstrate the implications of these design choices. The expected outcomes encompass actionable insights, validated through rigorous evaluations, and the development of an energy prediction tool for ML-based system development, to help in the decision-making process. This work contributes to the broader field of Green AI by addressing a critical gap and guiding the transition towards a more sustainable AI landscape.