Developed managed Dynamic Resource Allocation for LLMInferenceService within the opendatahub-io/kserve repository, enabling on-demand provisioning of accelerator devices such as GPUs and TPUs through annotation-driven resource requests. This approach eliminated the need for manual resource claim templates, reducing operational overhead and minimizing the risk of misconfiguration. Leveraging Go, Kubernetes, and cloud computing expertise, the solution aligned resource allocation with workload demand, improving cluster utilization and accelerating large language model inference workflows. The work focused on enhancing DevOps efficiency by automating resource management, allowing for more scalable and responsive deployment of LLM services in cloud-native environments.