One multimodal AI model for all your subsurface needs.

We pretrain SciLM Subsurface on vast amounts of physics simulation data, so you do not have to train it on your oil field's data and can start using it from day one.

GeologyWell logsSeismicOil productionWell locationsDevelopment planGeologyWell logsSeismicOil productionWell locationsDevelopment planSciLMSubsurface
Geology, wells and a plan in. Oil production out, in seconds.A geology in. Where to drill every well, and the plan to drill them.Wells, logs, seismic and production in. Every geology that fits, in minutes.Wells in. A geology that honors every one of them. Try it below.One AI model, every direction.

Try the demo of our current model.

ResFlow, the open model behind SciLM Subsurface, builds a reservoir from a depositional environment, an outline and wells, and honors every well you place. The production-conditioned inverse and the surrogate are training now.

Runs on Hugging Face, whose free GPUs give each visitor a few minutes a day. If a generation is refused, the pre-generated examples still work, or open the demo on Hugging Face and log in for more.

Status. Conditioning on wells runs today in ResFlow, open on GitHub with the dataset on Hugging Face. The surrogate ships in the first quarter of 2027; the inverse, with seismic conditioning, in the third.

If the training data or AI model you need does not exist yet, write to us.

Would you like to try SciLM Subsurface or SciLM Atoms on your own data? Let us know.