TGS slashes seismic AI model training from 6 months to 5 days using AWS SageMaker HyperPod with near-linear scaling

TGS deployed a Vision Transformer-based Seismic Foundation Model (SFM) on Amazon SageMaker HyperPod achieving near-linear scaling for distributed training

Training time reduced dramatically from 6 months to just 5 days through optimized distributed training infrastructure

Expanded context windows enable analysis of larger seismic volumes than previously possible, improving model capabilities

AWS SageMaker HyperPod provided the distributed training foundation necessary for handling massive seismic datasets at scale