Mission control for 50K solar sites.

Helios was flying blind across 50K sites. Outages were detected by customers calling in. The team needed real-time visibility and automated remediation.

1. 01Architected a Kafka pipeline for 1.2M events/min with replay.
2. 02Trained anomaly models on 18 months of historical telemetry.
3. 03Built a control plane in Go for one-click region failover.
4. 04Shipped operator dashboards with role-based runbooks.