Worked on the vllm-project/aibrix repository to automate synchronization between scheduling configuration and PodGroup state within a Kubernetes environment. Developed a backend feature in Go that ensures the PodGroup automatically updates whenever the RoleSet’s scheduling strategy changes, eliminating the need for manual intervention and reducing configuration drift. The implementation included targeted unit tests to verify that updates to the PodGroup occur correctly and to prevent regressions. This approach improved operational reliability by maintaining alignment between configuration and runtime state. The work demonstrated practical application of Go and Kubernetes controller patterns, focusing on robust state management and automated testing for backend systems.