A ten-week hands-on roadmap that teaches LLM inference serving by deploying, measuring, and optimizing one OpenAI-compatible service.