This document discusses using Azure Machine Learning and stream processing to enable predictive maintenance for aircraft engines. It describes a use case of predicting whether a device will fail within the next two weeks using real-time sensor data streams. It then outlines the challenges of stream processing and applying machine learning to streaming data. The proposed solution architecture involves using Event Hub for data ingestion, Stream Analytics for stream processing and aggregations, Machine Learning for model training and predictions, and DocumentDB for storing prediction results. It provides examples of the Stream Analytics and Machine Learning workflows used to enable predictive maintenance from real-time sensor data streams.