Solar atmosphere inversion

Utilized deep learning to solve inverse solar physics problem. Training models on over 100TB of simulation data. Model to be deployed in Daniel K. Inouye Solar Telescope workflow.

Open ocean SAR satellite foundation model

Used self-supervised representation learning to train a foundation model for SAR WV mode. Applied model to retrospectively analyze 9 years of global-coverage data.

Imaging-derived mortality marker from dual energy X-ray absorptiometry

Consolidated multiple longitudinal NIH studies with different imaging views to train a self-supervised DXA vision model and derived a multi-modal mortality marker based on body composition information in tabular fitness markers.

Physics-informed particle identification model

Designed neural network architecture based on known physics principles for Belle II detector Kaon/ Pion particle identification.