Back to jobs
Job Description
- Contribute to the development and maintenance of Tokamax, a unified open-source kernel library, creating a home for high-quality, well-tested, easy-to-use, and performant kernels available to both internal and external users.
- Build infrastructure and tooling for kernel development, including bench-marking suites, auto-tuning frameworks, performance analysis tools, debugging tools, and continuous integration pipelines to ensure the correctness and performance of custom kernels across different hardware and model configurations.
- Design, develop, and optimize high-performance custom kernels (using languages like Pallas, Mosaic, and Triton) aiming TPU and GPU architectures for key machine learning operations.
- Investigate and implement custom kernel support for new accelerator hardware generations/features and emerging ML operations.
- Contribute to the documentation and usability of kernel libraries tools and libraries to lower the barrier to entry for researchers and engineers looking to write or leverage custom kernels.