Worked on cross-framework interoperability and MLIR optimization, contributing to both google-ai-edge/ai-edge-torch and tensorflow/tensorflow repositories. Delivered unified interpolation attributes across JAX and PyTorch, standardizing NCHW handling and attribute naming to ensure consistent behavior between frameworks. Improved CI/CD reliability by re-enabling critical tests and updating dependencies, unblocking pipelines and enhancing testing stability. In TensorFlow, added an extensible MLIR dialect registry for TensorFlow Lite and streamlined optimization tooling by removing unused passes and dependencies, reducing build overhead. Leveraged C++, Python, and deep learning frameworks, focusing on build optimization, dependency management, and compiler design to improve scalability and developer productivity.