Context
Bachelor’s final-year internship at Oceantrade Madagascar, on an EV charging-station platform. Thesis topic: “Advanced monitoring and intelligent detection of fraud and anomalies in an electric vehicle charging station system”.
Problem
Collect and transform station data, characterise how stations operate and are used, then automatically flag atypical behaviour and anticipate failures.
Approach
- Backend services in Python/FastAPI and ETL pipelines to collect and transform station data.
- Processing orchestrated with Apache Airflow; inter-service messaging with RabbitMQ.
- Statistical analysis of OCPP data to characterise how stations operate and are used.
- Isolation Forest to detect anomalies and atypical behaviour in operational data.
- Random Forest to anticipate failures and guide maintenance interventions.
- Contribution to testing, debugging and deploying the platform.
Stack
- Python
- FastAPI
- ETL
- Apache Airflow
- RabbitMQ
- OCPP
- Isolation Forest
- Random Forest
Status
Completed (2025)
Internship completed in November 2025; bachelor’s thesis (degree awarded in 2025).
What I learned
An anomaly-detection model is only as good as the pipeline feeding it: without properly collected, transformed and orchestrated data, neither Isolation Forest nor Random Forest says anything useful.