• NASA SWOT foundation model: Leading end-to-end design and training of a domain-informed self-supervised foundation model on terabyte-scale satellite data using multi-GPU distributed training on HPC infrastructure.
  • Production satellite ML pipelines: Building image retrieval and semi-supervised learning systems on the WV-Net foundation model, processing millions of ESA Sentinel-1 SAR images for atmospheric science and oceanography applications.
  • Clinical foundation model: Refining a dual-energy X-ray foundation model with fairness-aware evaluation pipelines, ensuring stable subgroup performance across mortality risk, cardiovascular health, fracture risk, and metabolic disease prediction.