Design-partner beta · Free while we build with you

Drop in your fault sticks. Watch the agents reason.

FaultMind runs your interpretation geometry through four visible agents — corrected throw profiles, classified relay zones, bed-parallel-slip flags and a seal-risk screen. Every finding traceable to a published rule, with a confidence level and P10/P50/P90, exportable straight back to Petrel.

< 30 min to first real insight No seismic — two text files Anonymized by default Français / English
faultmind — Rhine Graben, Soultz demo
Judgment encoded from the literature — Peacock & Sanderson 1994 Yielding et al. 1997 Manzocchi et al. 1999 Walsh et al. 2003 Fossen & Rotevatn 2016 Allan 1989
Capabilities

Detection is everywhere. Judgment isn't.

AI pickers flood interpreters with fault sticks; simulators demand numbers nobody has time to derive. FaultMind sits between the two — the structural judgment layer.

Structural QC of the whole network

Nobody sanity-checks an assembled fault network — FaultMind does. Impossible intersections, orphan segments, throw discontinuities, one fault wrongly split in two: each flagged with a suggested fix, before it poisons your model.

Works on manual picks and Geoteric / PaleoScan / OpendTect exports

Relay intelligence

Relays identified and classified — bifurcating vs unconnected — with a breach probability, not a hunch.

P10 / P50 / P90 on every call

Corrected throw

Segment profiles under-report displacement where relays transfer it. FaultMind exports the aggregate profile your model actually needs.

Walsh et al. 2003

Bed-parallel-slip flags

Single-horizon throw anomalies against the growth-normalized trend — the signature restoration quietly absorbs unless someone flags it.

Tanner 1989 · Ferrill et al. 2017

Seal-risk screen

SGR with calibrated 15–20% thresholds, reservoir juxtaposition windows, and transmissibility multipliers ready for OpenFlow, tNavigator or Eclipse.

Yielding et al. 1997 · Manzocchi et al. 1999

Explainable by construction

Every finding is a card: claim, named rule, citation, confidence, uncertainty. Disagree in one click — your corrections train the rules.

Deterministic rules decide; agents narrate
How it works

From export to insight in one coffee.

Open a demo — or drop two files

Fault sticks + horizons (IESX, ASCII, CSV, XYZ, IRAP), the exports every interpreter already has. No SEG-Y, no gigabytes, no data-room clearance.

Anonymized before it lands

Coordinate shift and name hashing, on by default. We store unlocated geometry — nothing that says where your field is.

“It’s just unlocated geometry in a text file.”

Watch four agents reason

Ingest & QC → Geometry → Structural Intelligence → Uncertainty & Report, streaming live. Then export corrected sticks, fault-property tables and a PDF report — in French or English.

1.2 sto the first finding on the Soultz demo
10 rulesfrom the published literature, cited on every card
200Monte-Carlo realizations behind each P10/P50/P90
FR / ENinterface, reasoning log and report — day one
Design-partner beta

Free while we build it with you.

We're onboarding a first cohort of 6–8 teams — consultancies, operators with faulted assets, CO₂ storage and geothermal groups. No procurement, no payment details, a one-page agreement.

What you get

  • Unlimited runs on the public demo projects
  • 2 private projects (90 days), ~50 faults each
  • Full findings, exports and the PDF report (beta watermark)
  • A direct line to the builders — weekly build cadence

What we ask

  • A 30-minute feedback call every two weeks
  • Anonymized geometries usable for validation
  • On conversion: a named quote or logo

Who it fits

  • Fault-seal & compartmentalization studies
  • CO₂ storage permitting evidence
  • Fault-hosted geothermal exploration
Request beta access Beta reports are watermarked “not for regulatory use”. Regulatory-grade, versioned reports arrive with the paid tier.
Get beta access