Edge Preprocessing

Some validation and feature extraction runs at the edge, near the machine, to cut latency and volume before telemetry travels to the archive and twin.

Compute at the source

Not all processing can afford a round trip to the data center. The fabric runs a first pass of validation and feature extraction at the edge, on hardware close to the machine, so the fastest features are ready inside the control loop's budget and the data volume shipped upstream is reduced. This mirrors the edge-to-cloud latency gradient of the wider architecture.

What runs at the edge

• Anti-alias handling, HF filtering, and first-pass range and dead-sensor checks.
• Fast features the loop needs immediately: strain deltas, mode amplitudes, plug density.
• Compression before uplink, to fit the link's bandwidth.
• Timestamp de-embedding, so upstream data arrives already time-aligned.

Edge versus cloud split

The split is set by latency and information needs. Features required within the sub-10 microsecond control neighborhood are computed at the edge; heavy inversions like the full core pressure map and the offline multi-physics Monte Carlo run in the data center on L0. The parity contract guarantees a feature computed at the edge matches the same feature recomputed from the archive.

Consistency and governance

Edge preprocessing uses the same versioned feature and calibration definitions from the registry as the cloud path, so there is one definition, two execution sites. Every edge-computed value carries its lineage and quality score upstream. Edge nodes serve both the breeder and the burner and are a design specification for machines whose FOAK operation is expected near 2030.