In July 2025, Gabriel Chinora refactored the ONNX operator parsing framework in the ROCm/AMDMIGraphX repository, introducing dedicated builder classes for batchnorm, celu, einsum, gemm, and mean_variance_normalization operators. Using C++ and applying principles of code organization and refactoring, Gabriel centralized operator implementation details, reducing code duplication and improving maintainability. This work established a more modular architecture for ONNX parsing, making it easier to add and test new operators in the future. The changes addressed long-term extensibility and maintainability challenges, laying a foundation for faster ONNX operator support and more efficient compiler development workflows within the project.