Open-source LLMs work in demos. Production is different. This masterclass breaks down the full engineering stack required to run LLMs at scale: post-training, deployment control, tail latency management, and observability;illustrated with real production architectures and customer use cases.