Abstract: With the mounting pressure on organizations to modernize their infrastructure, particularly in regulated and mission-critical settings, there exists a rising demand of on-premises AI/ML automation pipelines that can provide enterprise-level security, compliance and control. This paper presents an overall engineering solution to create a fully automated AI/ML pipeline on-prem and using GitOps principles and self-hosted runners of GitHub Actions. The suggested pipeline combines infrastructure-as-code, container orchestration, and continuous integration/continuous delivery (CI/CD) pipelines to allow fast and repeatable models and supporting system deployment, and auditing it. The architecture enables data sovereignty, low latency and regulatory constraints by using self-hosted runners that operate on private infrastructure, which is important to sectors in which the national interest in infrastructure modernization may exist. The article describes the choice of tools chains, security enhancement, deployment methods, monitoring and governance habits. Use-case scenarios reveal the ways in which the pipeline contributes to the controlled environments and show how my work as the automation lead accelerated the creation of the secure AI ecosystems. The results show that on-prem pipelines constructed using GitOps and self-hosted runners have the potential to satisfy enterprise demands of repeatability, traceability, and resilience and decrease reliance on public cloud services.

Keywords: AI/ML pipeline, on premises infrastructure, GitOps, GitHub Actions, self hosted runners, CI/CD automation, infrastructure modernization, regulated environment, data sovereignty, secure AI ecosystems.

Downloads: | DOI: 10.17148/IJIREEICE.2024.12219