NV-Tesseract × Hullproof — complementary signal channels.

1. What NV-Tesseract is good at

NVIDIA's NV-Tesseract is a family of foundation models for time-series anomaly detection. Fed unified, AI-ready data from Cognite Data Fusion, it detects anomalies across diverse equipment types — wells, heat exchangers, pumps, compressors — by reading the sensor signal stream and flagging unusual patterns.

Aker BP's operationalization started with critical equipment safeguarding: catching anomalies in heat exchangers and well-stream sensors before they become unsafe or shut-down events. The model lives in CDF, the signal flows in from existing sensor infrastructure, and the alerts go to Atlas AI agents + the technical team.

2. What NV-Tesseract doesn't do

Foundation time-series models work on what the sensors see. They don't see asset coatings. They don't see biofouling extent. They don't see crack propagation at weld toes. They don't see subsea structure corrosion patterns. The physical-asset condition layer that drives long-term integrity-management decisions is visual — not time-series — and it sits outside what NV-Tesseract is built to read.

3. What Hullproof adds — same Atlas AI consumption layer

• Different signal channel. Hullproof reads footage (ROV, drone, handheld, fixed-camera, satellite, crawler, AUV). NV-Tesseract reads sensor time-series. Neither replaces the other.
• Same destination. Hullproof exports CoatingPassport to Cognite Data Fusion as Asset + Event records — the same shape NV-Tesseract anomalies use. Atlas AI agents consume both as a unified pane.
• Same audit trail. NV-Tesseract anomalies and Hullproof findings both carry source-frame / source-signal lineage. Defensible for EU AI Act high-risk-AI-system documentation, for PSA Norway audit, for class-survey input.
• Same agentic interface. Atlas AI agents call NV-Tesseract for sensor anomalies; they call Hullproof MCP tools (or the live VQA endpoint) for footage-based condition questions. One agent, two modalities, one Cognite-grounded answer.

4. The integration pattern

For Aker BP — and for any operator running Atlas AI + NV-Tesseract today — Hullproof plugs in at exactly the same layer:

1. Hullproof emits CoatingPassport from inspection footage (drydock, ROV campaign, drone sweep, fixed-camera mount).
2. CoatingPassport renders into CDF Asset + Event records via the live CDF exporter.
3. Atlas AI agents already consuming NV-Tesseract anomalies now also consume CoatingPassport findings — same Asset hierarchy, same Event taxonomy, same trust chain.
4. When a finding warrants action, Atlas AI orchestrates the response (work order, follow-up inspection, autonomous-robot dispatch via Hullproof's VLA layer if co-developed).

5. Why this matters for Aker BP specifically

Aker BP's public AI-first commitment is about autonomous operations across the asset base. NV-Tesseract closes the sensor-anomaly loop. The structural / coating / corrosion / biofouling loop on physical assets is the one that drives offshore integrity-management at NORSOK N-005 grade — and that loop runs on visual data, not sensor data. CoatingPassport is the structured intermediate that makes the visual-data loop agentic- readable in the same way NV-Tesseract makes the time- series loop agentic-readable.

For the Aker BP Eureka or equivalent program: Hullproof is the missing modality in the asset-condition data layer. Same destination, complementary channel.