Production .NET Core platform that ingests IoT energy data, runs ML.NET predictive models, and surfaces anomalies in real time.
Industrial clients had no unified way to monitor energy use across distributed smart meters, and manual reporting consumed significant operational hours each week.
Layered architecture with a dedicated ML service module. IoT ingestion via REST endpoints, time-series storage in SQL Server, ML.NET regression models for consumption forecasting, anomaly detection running on a scheduled background worker.
Chose ML.NET over a separate Python service to keep the prediction layer inside the .NET runtime — fewer moving parts in deployment, easier for the client's ops team to maintain. Used Angular lazy-loading to keep the dashboard fast under heavy chart loads.
I would extract the ML pipeline into its own service today and add proper observability around model drift — both gaps I now see clearly after a year of MSc study in AI.