Networking for AI inference model serving - GKE only and for all other backends