During August 2025, Marko Stojko developed core vision-related features for the tenstorrent/tt-metal repository, focusing on the introduction of the VisionEmbedding class to support vision model capabilities. He enhanced deployment reliability by refactoring code, expanding test coverage, and updating documentation, all implemented in Python with supporting YAML configuration. Marko improved CI pipelines and model configuration to ensure compatibility across multiple models and deployment environments, reducing integration risk and enabling faster iteration. Additionally, he updated the Tracy submodule to incorporate the latest features and fixes. His work demonstrated depth in deep learning, DevOps, and robust unit testing practices throughout the project.