Nagarjuna Devarayi
Master of Science in Data Science
Regis University
Expected Graduation: Spring 2026
Denver
Master of Science in Data Science
Regis University
Expected Graduation: Spring 2026
Denver
Nagarjuna Devarayi is a Master's student in Data Science at Regis University, bringing a strong technical foundation from his Bachelor's in Electrical and Electronics Engineering (EEE) at KKR & KSR Institute of Technology and Science, Guntur, Andhra Pradesh. With a passion for turning data into meaningful solutions, he is actively expanding his expertise in machine learning and AI, as demonstrated through his work on skin lesion segmentation and classification using deep learning. Beyond academics, Nagarjuna is an enthusiastic cricket and badminton player, and he aspires to independently develop his own application, driven by curiosity and a hands-on approach to technology.
This study presents an automated skin lesion segmentation and classification system trained on the ISIC 2019 dataset, comprising 25,331 dermoscopic images across 8 disease classes. A two-phase transfer learning approach using EfficientNetB2 was employed, with BCDU-Net performing lesion segmentation to reduce background noise before classification. To address a severe 54:1 class imbalance, four strategies were evaluated including class weighting, feature-space SMOTE, and conditional GANs (CGANs) with CGANs yielding the best macro F1 score of 0.79 by generating new pixel-level training data for rare classes. Grad-CAM heatmaps confirmed model interpretability by validating focus on lesion boundaries, highlighting the importance of using macro F1 and per-class recall over accuracy in medical AI contexts.
Feel free to reach out for collaboration opportunities, questions about my projects, or just to connect!
I'm always open to discussing new projects, creative ideas, or opportunities to be part of your vision.