During June 2025, Sihan Chen developed a performance optimization for the GPT-SoVITS model loading process in the opea-project/GenAIComps repository. By leveraging Git and Shell scripting, Sihan implemented a shallow clone strategy and removed the .git directory after download, directly addressing high memory usage during model loading. This approach reduced the runtime memory footprint and improved startup times, laying groundwork for future model-loading enhancements. The work was completed as a single feature, with all changes validated through continuous integration and testing. Sihan’s contribution demonstrated thoughtful use of Shell and Git to solve a targeted performance bottleneck in model deployment workflows.