Worked on the openvinotoolkit/openvino.genai repository to enhance the accuracy of power measurement during text generation tasks. Developed a feature in Python that instruments precise timestamps immediately before and after model.generate() calls within the text_generation.py module. This approach enables more reliable telemetry by capturing the true duration of generation workloads, supporting improved power and energy profiling. Additionally, the solution propagates a log prefix parameter to ensure consistent logging across text generation functions. Leveraging skills in Python programming, data analysis, and machine learning, this work facilitates more cost-aware optimization and benchmarking for generative AI applications without introducing any bug fixes.