workshop · Upcoming
SEG: Advancing Data Analytics & Machine Learning for Exploration Geophysics
SEG Research Committee's second workshop on machine learning for exploration geophysics. The first edition (Houston, May 2022) drew 80 attendees + 43 talks; 2026 expects materially larger.
SEG's Research Committee runs this workshop every ~4 years to take the technical pulse of the geophysics-AI community. The first edition (May 2022, Houston) had 80 attendees + 43 talks and produced one of the more cited workshop volumes in recent SEG history.
What's different in 2026
The four years since 2022 have been transformative for geophysics ML:
- Foundation models for seismic — GPU-bottleneck no longer the limiting factor; data is
- Self-supervised pretraining on multi-basin archives is now the default, not the exception
- Equivariant architectures for atomic-scale chemistry (MACE, Allegro) are starting to influence subsurface analogues
- Domain adaptation moved from academic curiosity to operator deployment
EarthScan participation
M Quamer Nassim + Tannistha Maiti are submitting talks. Topics under consideration:
- EAN-DDA at multi-basin scale — the F3 / Penobscot / Volve triple-transfer extension first shown at IMAGE 2025
- VeerNet as a foundation-model template — what we learned building a transformer-residual hybrid for raster log digitisation, applied to the broader "raster archive → queryable digital" problem class
Why this workshop matters
SEG workshops produce dense technical proceedings that the research community references for years. For EarthScan, participating here positions our work alongside the academic-research peer set we want to be measured against.
