Worked on the ZJUSCT/HPC101 repository to enhance debugging and observability for convolution validation in high-performance computing workflows. Developed a feature in C++ using CUDA that improves error reporting by identifying the precise indices—specifically n, output_h, output_w, and k—where mismatches occur during convolution output validation. This targeted diagnostic approach enables developers to quickly locate and address issues in numerical kernels, reducing the time spent on manual debugging. Although no major bugs were fixed during this period, the work focused on strengthening the reliability and maintainability of the codebase by providing more actionable information during validation failures.