Patented, in-production deep learning pipeline (DeepLabV3+ with mask-guided attention + Vision Transformer) for traumatic hemorrhage detection in 3D CT, deployed at the R Adams Cowley Shock Trauma Center.

Fine-grained VLM for anatomy-level CT understanding, achieving 0.86 mean F1 and 0.88 AUC across 95 injuries in the chest, abdomen, and pelvis (CLIP-style contrastive learning, anatomy pre-training).

Multimodal deep learning model predicting wildfire ignition risk from satellite imagery, weather, and grid data; fault-prediction pipeline for power infrastructure.

Deep CNN architectures for cooperative sensor perception in connected-vehicle (V2X) networks, with a multimodal LiDAR/camera/GPS dataset supported by Toyota.