In June 2025, Avik focused on improving error messaging for dynamic shape specialization in the pytorch/pytorch repository. He addressed a core bug by clarifying the use of Dim.DYNAMIC in error messages and suggesting actionable alternatives, which streamlined the debugging process for developers working with dynamic shapes. Avik updated multiple unit tests to ensure consistency with the new error handling logic, emphasizing robust debugging and error handling practices. His work, implemented in Python, reduced ambiguity in error reporting and improved the overall developer experience. The changes had cross-team impact, making dynamic shape debugging more efficient and accessible within the PyTorch ecosystem.