# Ray Serve deployments for GPU-shared AI inference