Akash Agrawal enhanced the pytorch/ao repository by developing a configurable extension to the Wanda Sparsifier, enabling per-layer observer configuration for model quantization workflows. Using Python and PyTorch, Akash implemented logic to attach observers to specific model layers based on an optional configuration, improving flexibility and maintainability for deep learning practitioners. The solution included robust unit tests to validate both custom configurations and fallback behavior when no configuration is provided, ensuring reliability across diverse usage scenarios. This work addressed the need for targeted sparsification and easier experimentation, resulting in a more adaptable quantization tool for machine learning model developers.