Geosteering with gamma rays.
ROGII Wellbore Geology
Year
2026
Role
Machine Learning · Geoscience
Overview
A Kaggle competition entry on wellbore geology prediction: estimating True Vertical Thickness (TVT) along horizontal wells from gamma-ray logs, so a drill bit can stay inside the oil-bearing layer.
Horizontal wells start vertical and turn to run through a thin layer. A gamma-ray sensor on the bit reads different values in shale and limestone, and type wells act as lookup tables from gamma-ray signal to TVT. The repo includes analysis notebooks and two tools: a map viewer for XY well layouts with 3D proximity and parallelism checks, and a geosteering simulator with gamma-ray correlation and multi-axis views.
The task
Predict TVT from gamma-ray readings, measured depth, coordinates and layer-depth markers (ANCC, ASTNU, ASTNL, EGFDU, EGFDL, BUDA).
Data quality
The repository documents 761 anomalous wells with non-standard layer counts, missing markers and irregular TVT changes.
Tooling
map_viewer.py and well_viewer.py for exploring wells and simulating geosteering.
