Worked on core modernization and reliability improvements across two open-source repositories over a two-month period. In Nixtla/statsforecast, modernized the C++ ARIMA implementation by updating data types, adopting a newer C++ standard, and introducing safer array handling with std::span, which improved numerical stability and maintainability for statistical modeling workflows. For netflix/metaflow, focused on enhancing asynchronous runner reliability by improving signal handling, error management, and inter-process communication through a temporary FIFO mechanism. Leveraged C++, Python, and asynchronous programming techniques to address edge-case risks and performance bottlenecks, delivering foundational improvements for more robust, maintainable, and stable data processing pipelines.