Worked on enhancing privacy-preserving machine learning capabilities in the APPFL/APPFL repository by integrating Opacus-based differential privacy into the training pipeline. Leveraged Python and PyTorch to implement configurable privacy options, including Gaussian mechanisms, enabling flexible differential privacy settings for federated learning workflows. Updated the ResNet model variant to ensure compatibility with Opacus, and extended the VanillaTrainer component to support privacy-preserving training. Additionally, maintained configuration management using YAML to streamline deployment and experimentation. Demonstrated the new features through an updated federated learning notebook, providing practical examples of differential privacy in action. Focused on robust, modular integration without introducing new bugs.