This research introduces an end-to-end pipeline for optimizing LLMs to reduce energy consumption while maintaining performance for practical applications.

• Implements quantization and pruning techniques to create energy-efficient LLM variants
• Demonstrates less than 5% performance drop while achieving significant energy savings
• Provides a practical framework for sustainable AI deployment in network support systems
• Evaluates models on real-world fault ticket analysis in communication networks

For engineering teams, this research offers immediate pathways to deploy powerful language models in resource-constrained environments, substantially reducing operational costs and environmental impact without sacrificing essential capabilities.

Energy-Aware LLMs: A step towards sustainable AI for downstream applications