Detecting hydrocarbons directly from seismic data represents one of exploration’s biggest challenges and opportunities. Subtle amplitude anomalies can signal potential reservoirs but confirming them requires analysing multiple seismic cubes simultaneously and accounting for geological context – a process that demands both expertise and computation.
Shell’s GeoCrawler, a multi-task deep learning framework, tackles this by detecting amplitude patterns, flat spots, and reservoir structures, boosting confidence in prospect evaluation. Yet, its large-scale models demanded High-Performance Computing (HPC) resources and coding expertise, limiting geoscientist access.
To solve this, Bluware deployed GeoCrawler on a scalable cloud platform using the VDS format, optimized for compressing and accessing large seismic data.
A custom connector links on-premise data to the cloud, integrating with geoscience tools. Bluware’s GPU-accelerated engine and Kubernetes clusters enable fast, parallel processing, cutting analysis time from weeks to hours.
The result is a robust end-to-end system where geoscientists can access AI-powered seismic interpretation directly – without deep coding skills or reliance on dedicated HPC clusters. The approach cuts processing time from weeks to hours and enables global collaboration, flexible scaling, and cost-efficient cloud deployment.
For AI developers, the framework also provides a faster path from research to production, allowing rapid iteration and deployment of new models with real-world data.
The Bluware–Shell collaboration, presented by Bluware’s Morten Ofstad at Dig X Subsurface 2025, demonstrates how operationalising AI is not only a technical achievement but a step toward making advanced hydrocarbon detection faster, scalable, and accessible for geoscientists worldwide.
Join us at Scandic Fornebu, December 03-04, 2025, to learn more. The program can be found on the conference website.