LG Uplus said on the 10th that at the 2026 Modern Agentic Applications Day hosted by Amazon Web Services (AWS), it presented a case study on building a platform to operate artificial intelligence (AI) models reliably in real-world services.
The event focused on automating infrastructure operations using Generative AI and assistant (agent) technologies. LG Uplus shared its experience building a platform that shifted from an on-premises-centered AI development environment to a flexible structure integrated with the cloud, and that efficiently connects the entire process from AI development to service operations.
The company said it strengthened the foundation to raise both the quality of AI services and operational stability.
The platform focuses on reducing the gap that used to occur between AI model development and service operations. Its core is integrating model training, evaluation, deployment, and operations into a single workflow.
LG Uplus adopted a hybrid infrastructure architecture based on Amazon EKS, AWS's managed Kubernetes service, to build the platform. It integrated its on-premises GPU infrastructure as hybrid nodes in an Amazon EKS cluster. It also said it reduced platform management overhead by running the Kubernetes control plane—which centrally governs the entire cluster—as a fully managed AWS service.
The company also said it improved how it uses graphics processing unit (GPU) resources. By applying a structure that allocates resources flexibly as needed, it reduced idle GPU time. It said it enhanced the system so GPU resources can be used efficiently when AI model training or service delivery is required.