Geothermal Prospectivity Engine
LithorAIAug 2026 - PresentCo-founder & technical lead
An engine that fuses multi-source open geoscience data (thermal, structural, seismic, satellite) into geothermal favorability maps with uncertainty ranges, built with faculty of the ITU Eurasia Institute of Earth Sciences.
Highlights
- Positive-unlabeled bagging with XGBoost, so unexplored ground is never treated as a confirmed negative
- Uncertainty maps and SHAP attributions for every target
- Validated with spatially blocked cross-validation and leave-one-field-out tests; caught a target-leakage flaw in the original prototype
