Fast model loading for AI inference on Amazon EKS

When you scale AI inference on Amazon EKS, every new pod must load model weights into GPU memory before serving traffic. We investigated where cold-start time goes and found two configuration-only changes to Run:ai Model Streamer that cut model startup time by 80-93% on subsequent launches, with no code changes.