- Enhanced image classification accuracy of CNN-based AI models from 70% to 93% using PyTorch.- Manually annotated 3,000 images as part of a comprehensive dataset creation initiative, resulting in a high-quality dataset of 5,000 images tailored for hazard prevention.- Leveraged extensive preprocessing techniques and model optimization strategies to boost neural network performance in real-world scenarios.- Engaged in collaborative research efforts, contributing to incremental enhancements through rigorous model validation and team-based experimental design.