Worked on the tenstorrent/tt-metal repository to deliver core vision-related features, including the introduction of a VisionEmbedding class with initial tests and supporting documentation. Leveraged Python, PyTorch, and YAML configuration to enhance model compatibility and deployment reliability, focusing on CI/CD improvements that support multiple models and environments. Updated the Tracy submodule to incorporate the latest features and fixes, ensuring up-to-date dependencies. Expanded unit test coverage and refined documentation to facilitate future development of vision-enabled models. The work emphasized robust testing, environment-aware configuration, and streamlined deployment, enabling faster experimentation and reducing integration risk for deep learning and machine learning workflows.