In May 2025, Zejun Chen developed Intel GPU (XPU) support for the unslothai/unsloth-zoo repository, enabling automatic detection and utilization of Intel GPUs alongside existing NVIDIA support. This work expanded hardware compatibility, allowing users on Intel-powered systems to leverage deep learning frameworks with improved GPU computing and hardware acceleration. Zejun focused on backend architecture to ensure seamless cross-vendor integration, laying the foundation for future multi-vendor GPU support. The implementation, written in Python, emphasized robust detection logic and maintainable code structure. Although no bugs were fixed during this period, the feature addressed a key adoption barrier for heterogeneous compute environments.