TPU ML Performance Engineer — JAX & Kernel Optimizer

Pageshift, a Research Lab in Zürich, is seeking an individual to optimize JAX workloads on TPUs, focusing on kernel development and end-to-end efficiency. The successful candidate will be responsible for implementing kernel-level optimizations, profiling, and validating improvements with data.

Applicants are expected to have a passion for storytelling along with a basic understanding of TPUs, JAX and XLA. Experience in writing Pallas kernels is a plus, further enhancing your candidacy.