Srdjan Djordjevic developed flexible tensor shape support for the ttnn.Tensor Pybind module in the tenstorrent/tt-metal repository, focusing on enhancing the Python-C++ integration. He replaced the previous hard-coded 4D shape constraint with a vector-based shape representation, allowing dynamic tensor shapes to be constructed at the Python-C++ boundary. This approach improved the module’s usability and reduced shape-related errors in downstream components, enabling broader integration with various models and pipelines. Srdjan utilized his expertise in bindings development, C++, and Python to deliver a targeted feature that addressed a specific limitation, demonstrating depth in cross-language API design and implementation.