Automating MLOps Pipelines with GitHub Actions and Azure: A Case Study on Diabetes Prediction

In recent years, there have been a lot of use cases for Machine Learning (ML) and Artificial Intelligence (AI), with use cases where companies try to operationalize on improving productivity, automating processes, and modernizing customer experiences. This paper aims to address the challenges of automating, rapid model deployment and Machine Learning Operations (MLOps) products and the operational bottlenecks in transitioning ML models from experimentation to production, such as inconsistent environments, lack of reproducibility, manual configurations, and limited CI/CD integration by building an end-to-end MLOps framework with a use case considering diabetes prediction model. MLOps addresses this by applying best practices and fostering a culture of efficient deployment. However, MLOps remains an evolving field with unclear implications for researchers and professionals. This paper presents a framework that leverages Azure cloud tools, i.e., Azure Machine Learning (AML), Terraform, and GitHub Actions, to automate the deployment pipeline, from dataset preparation to real-time API deployment. It also incorporates MLflow as an open source tool to streamline the ML lifecycle for experiment tracking and performance monitoring. Healthcare, particularly chronic illness prediction such as diabetes, provides a high-stakes and data-rich environment to validate the robustness of automated MLOps frameworks. This study uses the most widely used “Pima Indian Diabetes Datasets” prediction scenario, which offers structured, clinically relevant data. This framework shortens deployment time by 70%, improves reproducibility across environments, and achieves secure infrastructure management through environment variables and GitHub secrets, while approval gates and IAM roles ensure proper governance. The paper demonstrates how MLOps can transform ML workflows, enhance automation, scalability, and real-time decision-making in healthcare prediction.