Concept overview
This is a sample/demo entry used to test the portfolio layout, not a completed professional project. It explores how sensor time-series could be modelled for equipment health, using anomaly detection and remaining-useful-life estimation as the core prediction targets.
The intended stack is Python and PyTorch for modelling, with MLflow for experiment tracking so runs, parameters, and model versions stay comparable over iterations.
Intended serving direction
On the systems side, the concept outlines a FastAPI serving layer packaged with Docker and orchestrated through Kubernetes, with monitoring and drift detection to flag when a deployed model's inputs or predictions start to diverge from training conditions.
No performance metrics, measured results, or production deployments are claimed; this entry exists to demonstrate layout and content structure only.