Sector
Operational intelligence, document automation, and subsurface analytics for upstream, midstream, and refining operators.
Use Cases
- Well log and seismic interpretation using vision and transformer models for subsurface characterisation.
- Predictive maintenance on rotating equipment — compressors, turbines, ESPs — from vibration and SCADA telemetry.
- Production optimisation and virtual flow metering across well pads, refineries, and pipelines.
Subsurface to Refinery
Subsurface ambiguity, aging rotating equipment, and decade-old SCADA stacks make oil and gas the hardest place to ship AI — and the most lucrative when you do.
Where we work
Upstream operators with multi-basin assets, midstream pipelines balancing throughput against safety envelopes, and refineries running optimisation problems too large for spreadsheets and too sensitive for vendor black boxes. The data is heterogeneous, the integrations are decades deep, and the consequences of a wrong answer are physical.
We focus on three intervention points: the document layer (decades of well files, vendor specs, regulatory filings, P&IDs), the operational layer (SCADA, historian, MES integrations behind a safe inference boundary), and the engineering productivity layer (copilots for reservoir engineers, petrophysicists, and operations specialists).
Where AI changes the economics
- Document intelligence. Hundreds of thousands of unstructured documents — well logs, drilling reports, regulatory filings, vendor manuals — turned into a queryable knowledge surface for engineers. Hours to answers, not weeks.
- Predictive maintenance. Sensor-driven anomaly detection on critical rotating equipment, tied to maintenance scheduling and parts inventory. Downtime measured in dollars per minute; the math works.
- Production optimisation. Virtual flow metering, choke optimisation, and pad-level lift strategy informed by physics-aware ML — augmenting, not replacing, the reservoir engineer.
- Subsurface analytics. Seismic and log interpretation accelerated with vision transformers, with human-in-loop verification preserved as a hard requirement.
What we will not do
Run AI inside safety-critical control loops without HAZOP-grade review. Replace the engineer’s judgement on subsurface uncertainty. Promise autonomous operations of equipment that the manufacturer still ships with a manual.
What you get
A document intelligence platform or an operational AI deployment, instrumented for monitoring, drift, and audit. Built with your data, your integrations, your security posture, your safety culture. Not a vendor SaaS that ingests your data and disappears behind a logo.
Regulatory
- API 1164 and IEC 62443 for OT cybersecurity around control-system-adjacent AI.
- Process Safety Management — OSHA PSM, Seveso III — requiring HAZOP review and human-in-loop validation for any AI touching safety-critical loops.