Worked on the google-ai-edge/ai-edge-torch repository to enhance the reliability of the model export workflow for edge devices. Focused on improving error handling and file I/O by addressing a bug where exporting a model to a path without a directory could cause failures. The solution involved refining the logic to conditionally create directories only when a directory path was specified, and correcting the directory creation checks within the TfLiteModel export process. Using Python, the changes reduced export-time errors and improved the consistency of automated deployments, ultimately lowering maintenance overhead for distributing models to edge environments in production workflows.