Ray Serve makes scaling ML models feel easy — until you realize you’ve got data flying across clusters and no clue who’s talking to whom. This talk breaks down what it really takes to secure Ray Serve in Kubernetes: from in-cluster access controls to multi-cluster communication and secret management that won’t wake you up at 3 a.m. We’ll go through a real case study showing what worked, what failed spectacularly, and what we’d do differently next time. Expect some honest war stories, practical configs, and a few “don’t ever do this” moments