Worked on performance and quality improvements for the confident-ai/deepeval and NVIDIA/garak repositories, focusing on platform optimization and maintainability. Delivered Apple Silicon MPS backend support in BERTScorer, enabling Macs with MPS to leverage hardware acceleration for deep learning tasks. Addressed technical debt by removing a deprecated evaluator and its documentation to restore Python 3 compatibility and streamline maintenance. Enhanced test reliability in garak by adding targeted unit tests and refactoring probe module validation. Applied Python development, code refactoring, and test-driven development practices throughout, resulting in improved end-user performance, reduced maintenance overhead, and more robust software validation pipelines.