Daniel Wells

I work at the intersection of geophysics and machine learning, developing tools and models that bridge traditional scientific methods with modern AI. My geoscience background spans earthquake seismology, statistical seismology and earthquake early warning, ambient noise tomography, rare earth element characterization, and multiphysics simulation of serpentinization. I also have adapted to new domains in the form of cybersecurity anomaly detection, wireless signal intelligence and classification, and a variety of data science and machine learning and AI applications. Specifically, I have worked on surrogate modeling, physics informed machine learning, transformers, autoencoders, CNNs, GRUs, FCNs, LLMs w/RAG, and a variety of adapative, flexible, and general purpose models.

I hold a Ph.D. in Geophysics from the University of Utah and completed a Mendenhall Postdoctoral Fellowship at the USGS National Earthquake Information Center before joining Idaho National Laboratory. I am currently working at Utah State University as a research fellow, applying temporal point process modeling with GRU to forecast earthquake timing, magnitude, and location. I am also working on implementing static and dynamic Bayesian priors to improve inital location, timeliness, and magnitude calculations for earthquake early warning in low and noisy data regimes (i.e., when the first few seismometers record an earthquake and data is very sparse).