Real-world work spanning ML, systems, and full-stack development.
CNN-based classifier with pesticide suggestions; paper published in IJIRCCE.
Recommends similar movies with TF-IDF and cosine similarity; simple web UI.
Led a team building an ML model for automated defect triage and accuracy optimization.
Python-based blockchain auth with digital signatures & QR codes for real-time verification.
Linux kernel module in C to log syscalls and track processes via dmesg.
Merge Sort & QuickSort with OpenMP, MPI, and Pthreads; gprof profiling and automated CSV → plots.
Cross-platform expense tracking syncing to Google Sheets with admin reporting.
End-to-end CNN pipeline with data augmentation, class-balanced training, ROC-AUC evaluation, and Grad-CAM visualizations to highlight tumor regions. (Rebuilding after data loss.)
D4-equivariant CNN (0°, 90°, 180°, 270°) with e2cnn on UC Merced Land Use to improve robustness to rotations.
