Practicum Experience(s)
National Laboratory of the Rockies (2025)
Summary of Research
Scientific and engineering studies often depend on computer simulations or experiments that are too expensive to repeat thousands of times. Machine learning could help, but modern models typically need large datasets to perform well. My research develops generative models, which learn to produce new, realistic data, that remain accurate when only a small amount of high-quality data is available. A key idea is multi-fidelity learning: combining a few costly, accurate simulations with many cheaper, approximate ones so the model benefits from both. These methods can accelerate tasks such as design optimization and uncertainty quantification, making machine learning practical for scientific problems where data is scarce.
Annual Program Review Abstracts
Publications
Villatoro, C. J., Geraci, G., and Schiavazzi, D. E., "Assessing the Performance of Correlation-Based Multi-Fidelity Neural Emulators," Journal of Computational Physics, Vol. 565, 2026, p. 115170. doi: 10.1016/j.jcp.2026.115170.
Villatoro, C. J., Geraci, G., and Schiavazzi, D. E., "On Coordinate Encoding in Multifidelity Neural Networks," AIAA SciTech 2024 Forum, 2024.
Villatoro, C. J., Geraci, G., and Schiavazzi, D. E., "On Coordinate Encoding in Multi-Fidelity Neural Emulators," Tech. Rep. SAND2023-13916R, Computer Science Research Institute Summer Proceedings 2023, 2023.
Villatoro, C. J. and Van Zandt, N. R., "Impact of Non-Kolmogorov Turbulence on Scintillometer," 2021 IEEE Aerospace Conference (50100), 2021, pp. 1-7. doi: 10.1109/AERO50100.2021.9438206.
Awards
Science Graduate Fellow Award, University of Notre Dame, 2024
Gechtman/Osher Scholarship, Los Angeles Pierce College, 2017