Developed MindSpore model integration support for Ascend NPU devices within the kvcache-ai/sglang repository, enabling seamless end-to-end model execution on Ascend hardware. The work involved implementing distributed server initialization and detailed model configuration management to support scalable inference across multi-node environments. Comprehensive documentation and configuration guidance were provided to streamline onboarding and production deployment. Leveraging expertise in distributed systems, machine learning, and Python programming, the integration expanded the platform’s capabilities by allowing users to deploy MindSpore models efficiently on specialized hardware. This contribution focused on enhancing performance, scalability, and interoperability for machine learning workloads in distributed settings.