Production-oriented batch inference pipeline with FastAPI, Docker, model loading, validation, and structured prediction output.