Worked on stabilizing the PyTorch CoreML backend within the pytorch/pytorch repository by addressing a critical issue in the model conversion API. Focused on backend development and machine learning, the work involved adding support for a convert_to parameter and removing the requirement for a weights directory, which enabled more flexible model exports using coremltools. Utilized Python and PyTorch to resolve compatibility with the latest coremltools API, preventing export failures and improving deployment reliability for iOS. Enhanced maintainability and traceability by providing clear commit messaging and documentation, ensuring that future updates to the CoreML backend can be managed more efficiently.