My Vision Transformer faithfulness research received the Best Paper of Session Award at IEEE ESCI 2026. I built a TensorFlow evaluation pipeline to benchmark DAUC and SaCo metrics across 500+ images, quantifying a 2.0% faithfulness failure rate.
For my patented Brain Hemorrhage Detection System, I built a CNN-RNN hybrid model for binary detection and classification on CT scans. Systematic hyperparameter experiments helped improve accuracy from 90% to 95%.
On ZeroStratum, the national championship-winning fraud and security framework, I architected and tested a three-layer machine learning pipeline across 590K+ transactions. I optimized inference to 0.46ms and reduced operational costs by 75% while retaining 80% of security features.
