Developed end-to-end ATIF-native profiling support and DataFrame-based analytics for the NVIDIA/NeMo-Agent-Toolkit repository, unifying input formats to streamline downstream data processing. Leveraged Python and pandas to implement a workflow where both IntermediateStep and ATIF trajectory inputs are accepted, enabling seamless DataFrame generation for analytics and model training. Enhanced profiling outputs with detailed metadata and tracing, supporting per-tool provenance and backward compatibility. Comprehensive testing was conducted, including new ATIF/DataFrame tests and validation across existing profiler tests, ensuring robust integration. The work strengthened data analysis and visualization capabilities, allowing downstream modules to efficiently consume and process profiling data in standardized formats.