Adaptive Well Control with Reinforcement Learning
Train a controller to adjust well settings as new production measurements arrive, using reservoir simulations.
PythonPyTorchReinforcement LearningReservoir Simulation
papercode
Personal Projects
Reservoir Optimization & Control
Coordinated Multi-Well Control with MARL
Explore how wells that affect one another can coordinate their settings to improve shared outcomes in simulation.
PythonPyTorchMulti-Agent RLReservoir Simulation
papercode
Bayesian Optimization of Waterflood Operating Schedules
Search for better water-injection schedules while limiting the number of costly reservoir simulations.
PythonBoTorchBayesian OptimizationStreamlitPlotly
papercodedashboard
Subsurface Modeling & Scientific Machine Learning
Reliable Neural Surrogates for Porous-Media Flow
Build faster approximations of how fluids move through rock, and test when their predictions can be trusted.
PythonPyTorchNeural SurrogatesUncertainty Quantification
papercode
Conditional Generation of Geological Facies
Generate possible underground rock patterns that agree with simulated well observations in a synthetic benchmark.
PythonPyTorchDiffusion ModelsConditional Generation
papercode
Production Forecasting & Intervention Evaluation
Uncertainty-Aware Production and Produced-Water Forecasting
Forecast oil and water production, with ranges that show how uncertain each prediction is.
PythonTime-Series ForecastingUncertainty QuantificationStreamlitPlotly
papercodedashboard
Causal Evaluation of Well Reperforation
Study whether opening new flow paths in a well improves production, beyond changes that might happen anyway.
PythonCausal InferenceLongitudinal Data
papercode
Planned Methods & Tools
PythonUncertainty QuantificationCausal InferenceTime-Series ForecastingScientific Machine LearningPyTorchDiffusion ModelsBayesian OptimizationBoTorchReinforcement LearningMulti-Agent RLReservoir Simulation