Automating Cloud DevOps Workflows with Large Language Models: A Path to Self-Managed Infrastructure

Abstract

Traditional DevOps workflows in cloud computing involve manual scripting, infrastructure-as-code (IaC) complexities, and repetitive automation tasks. This paper proposes a novel LLM-driven DevOps automation framework, where natural language queries translate into executable cloud deployment scripts. The system utilizes LLMs to generate, validate, and optimize Terraform, Kubernetes, and CI/CD pipeline configurations, enabling faster, error-free infrastructure provisioning. Experiments show that the approach reduces deployment time by 45% and minimizes human errors in cloud automation, marking a step toward self-managed, AI-driven cloud operations.