SeisMind is an AI-powered desktop application that automates the seismic interpretation workflow — from well tie to reservoir characterization — using transformer-based machine learning trained on your own data. Your seismic. Your wells. Your models. Running on your machine.
13 specialized AI agents orchestrate the full G&G workflow end to end.
Your data never leaves your machine. All processing stays local.
Built by geoscientists — SEG-Y, LAS, ZMAP, Petrel ASCII. No conversion.
13 AI Agents. One Unified Pipeline.
SeisMind coordinates 13 domain-specific agents — ProjectManager, WellData, Seismic, SyntheticTie, Horizon, LogProcessing, SeismicAttribute, MLTraining, VolumeGeneration, Visualization, QC, Presentation, and Report — to execute complete reservoir characterization workflows with minimal manual intervention.
What used to take a team of three geoscientists several weeks now completes in hours.
Predict Well Logs from Seismic — Across Your Entire 3D Volume.
Train PyTorch transformer and mamba architectures on your well-to-seismic relationship. SeisMind learns the mapping between seismic attributes at well locations and measured log curves (GR, LLD, NPHI, RHOB, DT), then predicts those properties at every trace in your 3D survey.
The result: a spatially continuous reservoir model that honors your well data.
Automated Well-to-Seismic Correlation for Every Well in Your Project.
Computes acoustic impedance, reflectivity, and synthetic seismograms for each well, then cross-correlates with extracted seismic traces to find the optimal bulk time shift. Handles wavelet selection, sampling rate conversion, and TDR correction storage automatically.
Batch processing across 30+ wells in minutes with 8-panel diagnostic tie plots per well.
Your Data. Your Formats. No Conversion Required.
SEG-Y 3D seismic volumes, LAS well log curves, ZMAP horizon grids, Petrel ASCII well heads/deviations/tops/checkshots, and Excel/CSV tables. Import directly from Petrel, OpenWorks, or Kingdom project exports.
No intermediate format, no data loss.
See Everything — From Regional Map to Single Trace.
3D map view with deviated well paths, seismic trace display with horizon overlay, 5-track well log correlation sections with stratigraphic datum flattening, synthetic tie panels, and ML training dashboards with real-time metrics.
Built on PyQt6 and Matplotlib — publication-quality figures that export directly to PowerPoint and Word.
From Predicted Logs to Reservoir Properties — Automatically.
Derives VSH (shale volume), PHID (density porosity), PHIT (total porosity), PHIE (effective porosity), and SW (water saturation) at every trace. Each property exported as a full 3D SEG-Y volume.
Ready for import into Petrel, OpendTect, or any standard interpretation platform.
Your Seismic Data Never Leaves Your Machine.
Everything runs locally. SEG-Y files, LAS logs, well databases, trained models — and the agent orchestration itself — stay on your workstation. SeisMind exposes a local MCP server that your own AI coding agent drives; nothing is uploaded to Ameyem.
Compliant by design with data-handling policies at major E&P operators and national oil companies.
Load well heads, deviation surveys, checkshots, well logs, seismic volumes, and horizon grids from standard industry formats. SeisMind auto-detects Petrel exports and configures the project structure.
The SyntheticTie agent computes acoustic impedance, generates synthetic seismograms, and finds the optimal bulk time shift for each well. QC reports flag wells with poor correlation for manual review.
Select training wells, configure the transformer architecture, and launch training. SeisMind extracts seismic attributes at well locations, builds windowed training sequences, and trains PyTorch models with real-time validation metrics.
Apply trained models to predict log curves at every trace in your 3D survey — 466,000+ traces in a typical survey. Petrophysical volumes (porosity, saturation, shale volume) are derived automatically.
Generate well log correlation sections, map views, cross-plots, and facies distributions. Export to PowerPoint presentations and Word reports with publication-quality figures. Share SEG-Y prediction volumes with your team.
After installing, SeisMind runs a local MCP server, so a coding agent like Claude Code, Codex, or opencode can operate it for you in plain English. Your data never leaves your machine.
In Claude Code, one command registers SeisMind:
claude mcp add seismind -- seismind serveFor Codex, opencode, or any MCP-capable agent, add a server that runs seismind serve — for example, in an .mcp.json:
{
"mcpServers": {
"seismind": { "command": "seismind", "args": ["serve"] }
}
}Paste this to your agent:
“Download the Teapot Dome sample dataset, build a SeisMind project from it, and open it.”
SeisMind fetches the public Teapot Dome (US DOE / RMOTC) 3D survey and well logs, builds a complete project, and opens it. The first run downloads ~1.7 GB and is cached afterwards.
Then:
“Run a preliminary interpretation on the Teapot Dome project — give me a well-location map, a representative seismic section, and a summary of the survey.”
You get a basemap, a seismic section, and a survey summary in the project's interpretation folder. All of this works in the free edition — the same conversation scales into ML prediction, seismic attributes, and autonomous multi-well workflows on Pro.
The full desktop workstation, free for everyone.
For independents ready to interpret, not just view.
For interpreters who need seismic attributes at scale.
The full ML reservoir-characterization workstation.
For E&P companies and geoscience service firms.
All paid tiers are billed annually. Multi-seat volume discounts available — Enterprise offers custom pricing, SSO/LDAP, and on-premise deployment. Academic Pro licenses are free for verified researchers.
Let the agents do the heavy lifting.
SeisMind automates the repetitive workflows so you can focus on the geology. Download SeisMind Free, load a project, and run your first interpretation today.