Worked on the vllm-project/aibrix repository to implement batch storage integration using Volcano TOS, focusing on enhancing flexibility and scalability for batch job requests and outputs. Refactored the storage initialization process to support multiple storage backends, laying the groundwork for future integrations such as S3 or local storage. This approach reduced vendor lock-in and improved reliability for batch workflows. Leveraged Python for backend development and API integration, with an emphasis on cloud storage solutions. The work established a robust foundation for extensible storage architecture, enabling the project to adapt to evolving requirements and support diverse storage needs moving forward.