Worked on expanding the Conv2D testing framework within the tenstorrent/tt-metal repository, focusing on enhancing the robustness and reliability of the Conv2D implementation. Developed and integrated a comprehensive set of new test cases, including both passing and failing scenarios, to strengthen regression coverage and accelerate feedback during continuous integration. Utilized Python and deep learning techniques to ensure thorough validation of Conv2D features, supporting faster iteration and more reliable feature development. Improved documentation and reproducibility of the testing workflow, facilitating future maintenance and collaboration. The work emphasized systematic testing and contributed to sustainable quality assurance for machine learning acceleration features.