Worked on enhancing reliability and reproducibility in machine learning workflows using Python and concurrent programming. In the PriorLabs/tabpfn-extensions repository, implemented a deterministic Random Forest by standardizing random state usage, allowing users to define a single seed for consistent results and simplifying the codebase. Addressed a race condition in PriorLabs/TabPFN by introducing a file locking mechanism for concurrent model downloads, preventing file corruption and reducing bandwidth usage. Both changes were supported by expanded testing to ensure robustness under concurrent access. The work demonstrated a focus on backend development, software engineering best practices, and improving production deployment reliability through targeted improvements.