Worked on the DrewThomasson/ebook2audiobook repository to enhance the robustness of its multilingual processing pipeline. Focused on backend development using Python, the main contribution involved implementing a runtime guard that ensures Multi-Word Tokenization (MWT) in the Stanza NLP library is only applied to supported languages. This approach addressed a bug where unsupported languages could trigger runtime errors, thereby improving error handling and reducing maintenance overhead. By proactively checking language compatibility before invoking MWT, the update stabilized the pipeline for diverse language inputs and minimized error surfaces, reflecting a careful, detail-oriented approach to maintaining reliable NLP infrastructure.