How KoBold Metals accelerated discoveries through ML-driven exploration targeting
Models fuse geochem, geophysics and remote sensing to rank prospectivity. Drilling focuses on higher-probability zones, reducing time and cost.
Context
Early-stage exploration involved high uncertainty, with large areas requiring costly drilling. Traditional methods struggled to prioritize targets effectively and often produced low hit rates. The company needed a data-driven way to focus exploration spending.
Solution
ML models were trained on multi-modal datasets - geochemistry, geophysics, geology - to predict prospectivity across the region. The system generated ranked targets and highlighted areas most likely to contain mineralization.
Results
The approach accelerated discoveries and improved drilling hit rates compared with baseline methods. While exact financial outcomes were not disclosed, the project demonstrated meaningful value creation by reducing exploration waste and focusing drilling on higher-probability targets.
Data Inputs
Geochem, geophysics, hyperspectral, EM inversions, maps