Worked on the AI-Hypercomputer/tpu-recipes repository to enhance the command-line interface for serving large language models with VLLM. Focused on maintainability and usability, the work involved removing the deprecated --disable-log-requests flag from the vllm serve command, specifically targeting Llama-3.3-70B-Instruct and Qwen2.5-32B models. This change simplified the CLI, reduced potential configuration errors, and improved overall command clarity. The approach emphasized minimal, well-documented commits to streamline future contributions and onboarding. Leveraged DevOps practices and model serving expertise, with Markdown used for documentation, to ensure the repository remains reliable and accessible for ongoing machine learning development.