Wei Zhao developed a policy-driven deletion strategy for RayJobs in the ray-project/kuberay repository, focusing on improving resource cleanup based on job outcomes. He updated the Custom Resource Definitions (CRDs) and API documentation to clarify lifecycle semantics, and extended the controller logic to enforce new deletion policies during the RayJob lifecycle. Using Go and Kubernetes, Wei also implemented comprehensive tests to validate the behavior of these policies across both successful and failed job scenarios. This work enhanced lifecycle reliability and reduced the risk of orphaned resources, demonstrating depth in API design, CRD development, and robust testing within a Kubernetes environment.