Worked on the intel/ai-reference-models repository to enhance observability for inference workloads by improving model inference performance logging. Focused on refining CPU core detection within multiple run_model.sh scripts, the work enabled more accurate and reliable collection of performance metrics essential for diagnosing bottlenecks and optimizing resource usage. Leveraged bash scripting and Linux command line skills to standardize logging instrumentation across inference workflows, ensuring consistent and actionable data. Validated the end-to-end metric collection process using representative workloads, supporting data-driven optimization and more effective resource planning. The enhancements contributed to improved accuracy in performance monitoring and potential cost-efficiency for inference operations.