Worked on the modular/modular repository to deliver a machine learning accuracy testing workflow that integrates ml_dtypes for enhanced numerical operations. Established a Bazel-based build configuration for the accuracy_testing_framework, enabling reproducible builds and simplifying environment setup across development, CI, and production. Focused on standardizing dependency management and aligning build targets, which improved consistency and reliability of accuracy assessments for machine learning models. The approach accelerated model evaluation cycles and streamlined onboarding for new contributors. Utilized Python and Bazel to ensure compatibility and maintainability, resulting in a robust testing framework that supports cross-environment workflows and dependable model evaluation processes.