During March 2026, Zhao worked on the EvolvingLMMs-Lab/lmms-eval repository, developing a new performance metric for the Llava_OneVision1_5 class. Zhao introduced an fps attribute to enable frames-per-second based evaluation, supporting benchmarking and optimization of visual processing pipelines. The implementation leveraged Python’s object-oriented design and focused on performance instrumentation, ensuring compatibility and clear documentation within the codebase. This addition improved observability for real-time tasks, aiding data-driven capacity planning and SLA assessment. While no bugs were addressed during this period, Zhao’s work demonstrated depth in AI development and metric-driven engineering, contributing a targeted feature to enhance system performance evaluation.