Questions operators raise first.

Single-channel alarms fire when a parameter crosses a fixed limit. Sentinel classifies the developing multi-channel pattern earlier, identifies the mechanism, and attaches the supporting evidence — so the call is defensible in technical review rather than adding alarm noise.

Nothing on the rig. We’re cloud-based — we connect read-only to your existing rig-data hub (WITSML server, Pason DataHub, NOV / WellData, or your historian) over the network. No edge device, no new sensors, no crew change.

TLS 1.3 in transit, AES-256 at rest, per-well/per-role RBAC, tamper-evident hash-chained logs, and customer-selectable data residency (UK / US / EU). Cyber Essentials certified, hosted on AWS. The operator owns the data and can export it any time.

No. It’s physics-informed AI: physics constraints anchor what the model can conclude, the same inputs reproduce the same call (regression-gated against a frozen baseline), and every call carries a structured “why this call” trace.