About the project
Energy planning depends on reliable environmental data — but weather-station sensors drift, fail and produce unusual readings. This pipeline forecasts hourly environmental data and is engineered to catch real-time anomalies and physical sensor deviations as they happen.
What we built
- Automated data ingestion — scheduled collection of station data from APIs.
- Time-series storage — PostgreSQL with the TimescaleDB extension, built for high-volume time-stamped data.
- Feature engineering — time-series features that capture daily and seasonal patterns.
- Forecasting models — XGBoost and LSTM neural-network models for hourly predictions.
- Containerised deployment — the whole pipeline packaged with Docker so it runs the same everywhere.
Have a project like this?
From prototypes to production-ready systems — let's build it together.