Worked on enhancing hardware accelerator support in the antgroup/ant-ray and HabanaAI/vllm-hpu-extension repositories, focusing on documentation and reliability for Habana Processing Unit (HPU) resources. Improved Ray Serve and Ray Train documentation by adding detailed HPU resource descriptions, usage examples, and deployment configuration guidance, enabling users to manage HPUs alongside CPUs and GPUs. Addressed a correctness issue in bucket calculation logic within vllm-hpu-extension, ensuring accurate resource allocation. Leveraged Python, Markdown, and technical writing skills to clarify scaling and per-worker allocation patterns, reducing onboarding friction and configuration errors for teams deploying HPU-enabled workloads in distributed machine learning environments.