Worked on the GoogleCloudPlatform/magic-modules repository to enhance GPU workload management in Google Kubernetes Engine by introducing the gpudirect_strategy field to the node configuration. This feature enables users to explicitly control the GPUDirect strategy for GKE clusters, supporting performance tuning and cost efficiency for GPU workloads. The implementation involved updating the schema, integrating resource handling logic, and expanding documentation to provide clear usage guidance and examples. Leveraged skills in Golang, Terraform, and YAML to ensure the new configuration was fully supported across generated providers. The work focused on robust integration and clear documentation to improve user control over GPU performance.