Worked on the kvcache-ai/sglang repository to deliver a targeted enhancement to the benchmark serving command-line interface. Developed a new Python-based CLI argument that allows users to specify the served model name directly within the benchmarking script, streamlining model selection for evaluation tasks. This backend development effort focused on improving reproducibility and traceability of benchmarking results, enabling more efficient and flexible experimentation across different models. The approach emphasized commit-based development and careful CLI design, supporting faster setup and data-driven decision-making for model-serving workflows. No major bugs were addressed during this period, as the primary focus remained on feature delivery.