I'm a Computer Science graduate researcher at Georgia Southern University, working on 3D deep-learning pipelines for protein structure and actin filament segmentation from cryo-EM data. Before academia I spent two years building production backend systems in Go and Java. I care about clean, reusable code and research that holds up to scrutiny.
A research-driven path through academia and two years building production software.
Peer-reviewed work in deep learning, medical imaging, and bioinformatics.
From research frameworks to full-stack systems.
Hybrid CNN–Transformer for medical image segmentation extending TransUNet with dual-bias self-attention driven by fuzzy-cluster similarity and Gaussian spatial priors, plus an FCM-guided skip-gating mechanism that modulates CNN skip features to emphasize anatomically relevant regions.
Transformer–ResNet50 fusion architecture for classifying cracks, lack of penetration, porosity, and defect-free weld radiographs. Guided filtering and CLAHE preprocessing pushed test accuracy to 98.93% on the RIAWELC dataset.