Worked on the Cambridge-ICCS/FTorch repository to enhance inference efficiency in deep learning workflows. Focused on optimizing the ResNet inference example by introducing a torch.no_grad() context, which disables gradient computation during prediction. This approach reduced memory usage and improved performance, making the model more suitable for deployment in resource-constrained environments. The change was implemented in Python and leveraged deep learning best practices for inference optimization. Delivered a clean, production-ready update with minimal code changes, ensuring maintainability and ease of integration. No bug fixes were recorded during this period, with efforts concentrated on feature development and inference performance improvements.