Cloud Platform · 2025
Helio Grid for Helio Energy
A real-time control plane for 40,000 distributed solar assets across three continents.
- 11s
- Fault detection
- 40k
- Assets live
- 3
- Tools retired
The challenge
Operators were reconciling telemetry across four legacy dashboards, and a fault could take nineteen minutes to surface. Every minute of delay cost measurable generation revenue.
What we did
Replaced polling with an event-driven ingest pipeline on Kafka, collapsing four dashboards into one control plane.
Modelled asset hierarchy in PostgreSQL with time-partitioned telemetry and continuous aggregates.
Built a WebGL site map that renders 40k live nodes at 60fps using instanced geometry.
Shipped an incident timeline that replays any fault window frame by frame.
The outcome
Fault detection dropped from nineteen minutes to under eleven seconds, and the operations team retired three internal tools in the first quarter.