Industrial Failure AI & Dashboard

I am building an end-to-end solution that automatically detects mechanical failures on the factory floor and then presents the equipment-health status in a live dashboard.

The first part is a deep-learning model written in Python that ingests vibration, temperature and other on-board sensor streams together with image and video feeds from our existing cameras. The model should learn normal operating signatures, flag deviations that point to impending mechanical problems, and expose its results through an API. Feel free to rely on PyTorch, TensorFlow or any framework you are comfortable with—as long as the final code is well-structured, fully reproducible and comes with a short training/inference guide.

The second part is a lightweight monitoring layer in Power BI combined with a simple Power App. The dashboard must pull the model’s output in near real time, visualise key indicators such as anomaly scores or risk level, and let operators acknowledge alerts from a tablet on the shop floor.

Deliverables
• Trained detection model with source code and saved weights
• Inference pipeline (script or microservice) that pushes results to the dashboard’s data source
• Power BI report and embedded Power App with basic CRUD for alert status
• Brief documentation explaining data requirements, deployment steps and how to retrain

When you reply, focus on your direct experience with similar industrial analytics, especially projects that mix sensor data with computer vision and integrate with Power BI or Power Apps.