Problem:

A client in industrial manufacturing needs software that reads sensor data from manufacturing processes, predicts product quality, and recommends optimal process configurations.

Solution:

Historical batches are transferred to an Azure data lake, while ongoing data is streamed to an Event Hub.

In the development lab, data scientists develop an ML model training pipeline, which is synced with the build and update pipeline in Azure ML.

ML models are built and regularly updated with the latest data and released to the production pipeline. Predictions and recommendations are stored in a PostgreSQL database.

Backend and frontend are deployed on a virtual machine.