Build AI-ready APIs using REST, streaming, and async patterns that handle long-running model inference without timing out.