Research & Impact
Scalable solutions for energy, data, and control.
WindIO: Industry Data Standard
Led the development of WindIO, a machine-readable ontology adopted by five major software platforms (including NREL's FLORIS and DTU's PyWake). This work standardized the data exchange for complex wind energy systems, enabling interoperability across the entire modeling ecosystem and eliminating data compatibility bottlenecks.
Risk-Aware Control Strategies
Pioneered optimization under uncertainty approaches for wind farm control that reduced extreme mechanical loads by 47%. This work has been cited in over 100 research papers and referenced in industrial patents, becoming a standard methodology for robust "wake steering" control strategies deployed in the field.
Scalable Stochastic Gradient Descent
Developed a novel SGD algorithm capable of optimizing layouts for fleets of over 1,200 autonomous units. This approach enforces deterministic constraints within a stochastic framework, reducing computation time by 95% compared to legacy methods and enabling the design of gigawatt-scale energy systems.
Adversarial Sensor Errors for Robust Control
Addressed the "Arms Race" problem in adversarial learning by implementing a "Self-Play" training architecture. Trained control agents to survive active sensor corruption, resulting in systems that maintain high performance in both clean and hostile environments without the typical "alignment tax."
Safe Operation Envelope Prediction
Created a predictive validation tool that increased scenario coverage by 72%. By quantifying epistemic vs. aleatoric uncertainty, this framework focuses expensive validation campaigns only on high-risk operational conditions, acting as a force multiplier for safety verification.
