DF-01Foundational
Field Document OCR Pipeline
Turns decades of scanned reports, logs and maps into searchable text, including Persian and Arabic documents and handwritten annotations.
Data FoundationAI
Upstream Intelligence Platform
Field data, turned into decisions.On your infrastructure.
An on-premise platform bringing industrial IoT, data engineering and applied AI to upstream operations. 83 capabilities across 8 product families. Your data never leaves the field.
Scroll
Why upstream AI programmes stall
- 01
The data exists, and it is not usable
Forty years of reports sit in cupboards as scans. The historian compressed its tags to a dead band. Four systems give four answers for one well’s depth. None of this is visible until a project needs it.
- 02
Models fail without a foundation
A model trained on allocated rates learns the allocation, not the reservoir. Most published failures in this industry were data failures reported as algorithm failures.
- 03
Cloud platforms are unavailable or non-compliant here
Global platforms assume an outbound route and a licence server. In the markets we serve, operator data is required to stay on operator infrastructure, and frequently there is no route at all.
Six layers, two spines
Each layer does one thing and can be inspected on its own. Security and governance run through all of them. Open the architecture explorer to see what enters and leaves each layer.
- L5ExperienceThe dashboards, the field app, the alerts and the operations centre — where people actually meet the platform.
- L4Domain ServicesThe petroleum engineering logic — nodal analysis, allocation, reliability, integrity — as services anything can call.
- L3Analytics & ModelsWhere the physics models and the machine-learning models actually run, and where they are checked.
- L2Data FoundationWhere everything is stored, cleaned, named consistently and made traceable.
- L1Acquisition & Control IntegrationThe one-way road that brings control-system data out safely, and never lets anything back in.
- L0Field & EdgeThe sensors, cameras and small computers that sit at the wellhead and in the plant.
8 product families, 83 capabilities
You start with two and grow. Nothing here is sold as a monolith.
Drilling Intelligence
Reading the rig’s own sensors, second by second, to drill faster and stop trouble before it costs a day.
15 capabilitiesSubsurface & Reservoir Intelligence
Turning logs, cores, seismic and production history into a reservoir picture you can argue with — and act on.
15 capabilitiesProduction & Artificial Lift
Keeping every well producing at its best rate, and knowing which well to touch next.
14 capabilitiesFacilities, Rotating Equipment & Integrity
Hearing a machine fail weeks before it does, and inspecting the pipe that actually needs it.
8 capabilitiesHSE, Emissions & Security
Cameras and sensors that watch the things a person cannot watch for eight hours straight.
8 capabilitiesData Foundation
The unglamorous work that decides whether everything above it succeeds.
9 capabilitiesKnowledge & Copilot
Asking your own archive a question in your own language, and getting an answer with the page number.
6 capabilitiesOperations Digitalization
Replacing the clipboard, the spreadsheet and the phone call with something that works without signal.
8 capabilities
Three places operators usually start
- PR-01Foundational
Production Forecasting
Forecasts what each well will produce next month, and tells you which wells it is not confident about.
Production & Artificial LiftAI - HS-01Foundational
Field Safety Computer Vision
Uses the cameras already installed to spot missing protective equipment, people in restricted zones and unsafe proximity to lifting operations.
HSE, Emissions & SecurityAIIoT
How we deliver
Four stages, each with an exit criterion written before it begins. The first one frequently concludes that you should not buy the second.
- 01
Assessment
4–6 weeks
We find out what your data can actually support before either of us commits to anything.
- 02
Pilot
3–4 months
Two capabilities, one asset, a measurable result and a defined way to fail.
- 03
Deployment
6–12 months
Production hardening, integration with the systems that create work, and handover to people who will still be here in three years.
- 04
Scale
Ongoing
More capabilities, more assets, and a measured account of which ones are earning their place.
On-premise, and air-gapped where you need it
No licence callback. No telemetry. No model download at runtime. Updates arrive as signed offline bundles. The platform is designed to run in a field with no outbound route, and that property is tested rather than asserted.
- no licence callback
- no telemetry
- no runtime model download
- signed offline update bundles
- IEC 62443 zones and conduits
The platform in numbers
0Named capabilities
0Product families
0Delivery stages
0Standards supported
Recent writing
Most digital programmes fail before the first model is trained
The failure is almost never the algorithm. It is that the history the algorithm needed was never recorded.
On-premise is an architecture, not a compromise
Building for air-gapped deployment from the start produces a different system than porting a cloud product inward.
What early kick detection actually buys you
Minutes, not miracles — and the size of the influx you have to circulate out is roughly proportional to them.
Start with the assessment
Four to six weeks. We inventory your data, score each capability against what it actually requires, and tell you what is achievable now — including where the answer is “not yet”.