Energy-Aware Green Software Engineering: Optimizing AI Applications for Sustainable Computing Environments

Abstract

ABSTRACT The rapid proliferation of artificial intelligence (AI) and deep learning (DL) models across cloud, edge, and high-performance computing infrastructures has triggered an unprecedented surge in computational demand and energy consumption. As state-of-the-art machine learning architectures continuously expand in parameter size, the associated operational carbon footprint poses severe environmental and economic challenges. Green Software Engineering (GSE) has emerged as an essential discipline aimed at embedding energy awareness into every phase of the software development lifecycle (SDLC). This review paper provides a thorough, technically detailed investigation into energy-aware software engineering paradigms specifically tailored for artificial intelligence workloads. We analyze energy-bound hardware profiling, algorithmic efficiency optimization, neural network compression techniques (quantization, pruning, knowledge distillation), dynamic frequency scaling, and carbon-aware runtime scheduling. Furthermore, we synthesize empirical performance metrics, analyze accuracy-energy trade-offs, identify critical research gaps in current software metrics, and present a structured framework for sustainable AI engineering. Our comprehensive review establishes a roadmap for developers, system architects, and researchers to optimize AI applications for sustainable, energy-efficient computing environments without compromising prediction fidelity or functional utility.

KEYWORDS: Green Software Engineering, Energy-Aware Computing, Sustainable AI, Deep Learning Efficiency, Model Compression, Carbon Footprint, Dynamic Energy Scaling, Sustainable Computing Environments.