Case study
Predictive maintenance data pipeline
A reproducible pipeline that ingests simulated telemetry, validates time-series data, and flags maintenance anomalies.
What I am building
This project handles the data path behind an equipment monitoring product. It ingests simulated telemetry, validates and stores time-series records, and flags anomalies tied to maintenance thresholds.
The output is designed to serve more than one consumer. Scheduled jobs produce reporting-ready data that can feed dashboards, alerts, and downstream services without each client rebuilding the same rules.
Why it is separate
Keeping the pipeline separate from the interface makes its failure modes easier to see. Validation, delayed readings, duplicate events, and threshold logic belong to an explicit data contract rather than being hidden inside chart code.
Technical shape
Python and FastAPI provide the service layer, with PostgreSQL for storage. Docker keeps the development environment reproducible, and Terraform describes the AWS infrastructure rather than relying on a hand-configured deployment.
Current work
The project is in progress. The next pass will make retries and rejected records observable, document the ingestion contract, and test scheduled processing against larger simulated workloads.