Worked on the huggingface/diffusers repository to address reliability issues in Qwen image processing pipelines, specifically targeting the handling of negative prompt embeddings and mask management in edit and inpaint workflows. Using Python and leveraging deep learning and image processing expertise, introduced safeguards to ensure safe behavior when masks are missing, reducing runtime errors and improving user diagnostics. Enhanced the codebase by adding targeted tests for classifier-free guidance (CFG) scale behavior without negative prompt masks and improved logging to provide clearer warnings about prompt and mask issues. Updated documentation and comments to align with these changes, supporting safer and more predictable model usage.