IoT telemetry platform for 12,800 industrial robots across 94 facilities
Building an edge data pipeline that stays consistent even when factory network links go dark for hours.
Robot telemetry could predict bearing failures 3–5 days out — but only if it arrived. Factory networks dropped for minutes to hours several times a week, and the existing collector silently discarded data during every outage. Robots across three hardware generations also emitted incompatible schemas the central platform couldn't handle.
A local gateway in each facility buffers telemetry on-disk via a write-ahead log and replays events in order on reconnect, using sequence numbers the robots already emitted. Schema normalisation runs at the edge with OTA-updatable configs. Failure prediction models are deployed as ONNX runtimes on the gateway — predictions work even when the cloud link is down.
Unplanned downtime dropped 63% in the first year. 4.1TB of telemetry processed daily with no production data loss. Three previously unknown failure modes were identified from the now-complete dataset.
Have a Similar
Challenge?
Let's discuss how we can help you achieve similar results.
Start a Conversation