Developed the NeutronQuantizer for the NXP backend in the pytorch/executorch repository, focusing on enhancing model quantization for edge deployment. The solution introduced an annotation-based approach to prepare models, specifically targeting Conv2d and Linear patterns, which enables more efficient quantization and reduces model size while improving inference speed on NXP hardware. Comprehensive testing was implemented across multiple layers and operations to ensure the reliability and robustness of the quantization path. The work leveraged Python, PyTorch, and backend development skills, delivering a feature that streamlines the deployment of machine learning models on resource-constrained devices without addressing bug fixes during this period.