IoT-Exponenta
Anomaly detection over 2,000,000+ telemetry records from ~15,000 devices · ~92% accuracy · −40% manual QA.
- Built with
- Python, ETL
- Year
- 2024–2026
- Status
- Archived
I spent a year and a half at IoT-Exponenta as a data and automation engineer, working on industrial IoT and smart metering.
The main piece was anomaly detection over streaming sensor data — catching faulty devices before they turned into customer-facing failures, at around 92% detection accuracy, which cut manual data checking by roughly 40%. Feeding it meant turning over 2,000,000 raw telemetry records from about 15,000 metering devices into analysis-ready datasets with Python ETL.
The rest was the unglamorous work that makes the first part possible: redesigning how data came in and got triaged — validation rules, automated flagging, prioritised queues — which took about 30% off incident resolution time, and automating recurring processing that had been eating around ten hours a week. I coordinated across hardware, firmware and data in a team of five.