Projects

My research focuses on bridging the gap between highly robust mathematical AI models and lightweight, system-level software engineering for clinical deployment.

Submitted: MICCAI 2026 | System Architecture | Transformer Optimization

Meta-D: Metadata-Driven Attention via Tₘₐₓ Routing

Engineered a deterministic routing framework for multi-modal brain tumor segmentation. By utilizing a Transformer Maximizer Tₘₐₓ to calculate entropy, the model actively severs attention to missing or corrupted data streams. This architectural shift achieved state-of-the-art accuracy while shedding 24.1% of model parameters and dropping computational complexity to O(N).
Read Full Article →
Medical AI | Computer Vision | Live Web Deployment

Context-Aware 2.5D MRI Plane Classification

Solved the "near-skull" geometric ambiguity problem in MRI scans. Rather than utilizing heavy 3D volumetric networks, I architected a lightweight 2.5D Context-Aware Classifier that samples adjacent slices to learn local anatomical flow. This corrected metadata was gated into a tumor detection pipeline, reducing clinical misdiagnoses by 33.3%.
Read Full Article →
Accepted: IEEE ISBI 2025 | Uncertainty Quantification | Bayesian AI

Melanoma Detection via Uncertainty Quantification

Architected a 2D Bayesian uncertainty pipeline designed for mission-critical Out-of-Distribution (OOD) detection in dermatology. By dynamically filtering predictions based on low-confidence entropy scores, the model successfully slashed critical false-negative diagnostic failures by 40.5%.
Read Full Article →