Worked on the openvinotoolkit/openvino_tokenizers repository to enhance type safety within TensorFlow conversion extensions. Focused on refactoring the API for tensorflow::ConversionExtension, the developer improved the clarity and maintainability of the TensorFlow conversion workflow. By refining type definitions and conversion logic, the changes reduced runtime risks and minimized potential integration errors when handling TensorFlow operations. The work was implemented using C++ and CMake, leveraging OpenVINO and TensorFlow to ensure robust interoperability. This targeted feature addressed a specific pain point in the conversion path, laying the groundwork for easier future maintenance and more reliable extension of the OpenVINO tokenizers library.