From Prediction Models to MLOps: A New Era for Geoscience
- Kaveh Dehghan
- Apr 25, 2025
- 2 min read
Earlier this March, our co‑founder Kaveh Dehghan took the stage at EAGE Digital 2025 to present a forward‑looking perspective on how MLOps is reshaping subsurface workflows. His talk—delivered during the From Model to Life: AI Revolutionizing Field Operations session—highlighted the shift from static, infrequently updated models toward continuously learning, operationally agile subsurface systems.
For decades, reservoir modeling has depended on complex pre‑stack data, manual workflows, and long turnaround times. But geoscience is now entering a new phase: one where models evolve as fast as the data itself.
Why MLOps Matters for Subsurface Teams
Kaveh’s presentation explored how MLOps introduces automation, consistency, and continuous learning across the entire lifecycle of subsurface prediction. Key capabilities include:
Real‑time model updates that incorporate new wells, interpretations, and seismic refinements instantly
Lower data requirements, thanks to high‑confidence machine‑learning workflows built directly from post‑stack seismic
Automated pipelines that streamline data ingestion, quality control, model building, and output generation
Seamless integration of new interpretations without rerunning entire projects
Live models = live insights, enabling faster and more confident operational decisions
A North Sea Case Study: Continuous Learning in Action
One highlight of the session was a case study from the F3 North Sea block, showing how porosity predictions steadily improve as the model ingests new datasets and interpretations over time. The result is a dynamic, self‑updating system that reflects current subsurface understanding—an early example of agentic AI applied directly to geoscience workflows.
The Bigger Picture
This evolution is more than digital transformation. It is a fundamental upgrade to operational agility in subsurface decision‑making. By enabling models that adapt continuously, MLOps empowers teams to respond faster, reduce uncertainty, and scale insights across assets.
Want to bring continuous‑learning subsurface models into your workflow?
Get in touch with us at LithoSight—we’d be happy to discuss how MLOps‑driven geoscience can accelerate your projects.
