I need a Prognostic Health Management application that can learn from vibration data collected on ball bearings and warn me, through an interactive dashboard, when a failure is approaching. The core of the job is to design the end-to-end workflow: ingest raw vibration signals, clean and transform them (FFT, envelope analysis, time-frequency techniques—whatever best extracts degradation features), train a predictive model, then expose Remaining Useful Life or probability-of-failure metrics in an intuitive web dashboard. Vibration data will be the only input in the first release, so your signal-processing and machine-learning choices must squeeze maximum insight from that single source. I’m open to Python (NumPy, SciPy, scikit-learn, TensorFlow) or MATLAB toolchains as long as the final product is easy for me to retrain with new runs. Deliverables • Source code with clear comments and a short setup guide • A lightweight dashboard (Streamlit, Dash, or similar) showing live health indicators, trend plots and a simple traffic-light status • A sample dataset and step-by-step notebook that reproduces your results • Brief report explaining feature extraction, model selection and validation results Acceptance criteria: the model should detect at least 90 % of seeded failure events from the sample set with under 10 % false alarms, and the dashboard must refresh in near-real-time when new CSV files are dropped into a watch folder. If you’ve tackled bearing diagnostics before, particularly with vibration-only data