Developed and integrated RF-DETR, a new object detection and instance segmentation model, into the huggingface/transformers repository. This work involved implementing RfDetrForInstanceSegmentation and expanding the framework with modular configuration management, scalable execution, and support for configurable loss functions. The developer enhanced documentation by updating model cards, docstrings, and toctree entries, improving onboarding and maintainability. Extensive refactoring removed unnecessary dependencies, enabling distributed training and future scalability. Testing infrastructure was improved by migrating to dedicated tests and adding CPU device expectations. The project leveraged Python, PyTorch, and deep learning techniques to strengthen detection and segmentation capabilities within the repository.