Developed and integrated SVM kernel methods along with One-vs-Rest multi-class support into the OpenHUTB/nn repository, expanding the classifier’s capabilities to handle both linear and non-linear datasets with improved accuracy. The implementation maintained backward compatibility with existing linear SVM workflows, ensuring seamless adoption for current users. Leveraging Python, NumPy, and scikit-learn, the work focused on enhancing model versatility and scalability for production machine learning benchmarks. All changes were traceable through linked commits and issues, reflecting a disciplined approach to maintainability and auditability. This contribution established a robust foundation for further exploration of advanced kernel-based classification techniques.