Worked on performance and scalability improvements across huggingface/swift-transformers and ml-explore/mlx-swift-examples, focusing on AI model optimization and maintainability. Enhanced chat template rendering by introducing lazy memoization for applyChatTemplate and upgrading Jinja templating to version 1.3.0, reducing latency and improving stability. In the mlx-swift-examples repository, implemented a quantized cache path for the GPTOSS attention mechanism, optimizing memory and compute efficiency. Upgraded swift-transformers to support enhanced prompt handling, including video prompts, for SmolVLMProcessor. Leveraged Swift, Jinja templating, and caching strategies to deliver three new features that improved performance, reduced costs, and streamlined dependency management within one month.