Worked on the HazyResearch/ThunderKittens repository to refactor the kernel binding infrastructure, focusing on improving reliability and scalability for complex kernel integration. Leveraged C++, CUDA, and advanced template metaprogramming techniques to transform the existing utilities into a generic, type-safe solution. Introduced new abstractions, including bind_kernel and bind_function, which streamline the binding process and reduce repetitive boilerplate throughout the codebase. The approach utilized variadic templates and Pybind11 to enable flexible and maintainable kernel bindings. This work addressed the need for scalable integration patterns and laid the groundwork for future extensibility, with no major bugs reported or fixed during the period.