Year: 2026
Type: article
Abstract: Automating the path from a source repository to running cloud infrastructure remains a manual, error-prone task for many development teams. Recent work applies large language models (LLMs) and autonomous agents to this problem, spanning GitHub workflow generation, Infrastructure-as-Code (IaC) synthesis, drift reconciliation, and policy-bounded deployment decisions on cloud platforms such as Amazon Web Services (AWS). This paper reviews 30 recent publications at the intersection of artificial intelligence, GitHub-based automation, AWS, and automated deployment, with an in-depth analysis of eight core studies. We compare their architectures, AI techniques, evaluation methods, and reported outcomes, and organize the corpus into five themes. The review finds that LLM-based IaC generation reaches high deployment success mainly through iterative feedback, that generated infrastructure tends to over-provision relative to expert baselines, and that rigorous, comparable evaluation on realistic AWS workloads remains scarce. We conclude with open challenges and a conceptual roadmap for future work. The paper is literature-based and presents no system implementation.
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7235570689
