Build compilers and tools that efficiently map ML models (with a particular focus on always-on computing use cases) to the hardware ISA. Propose and implement fixes to the compiler to address these issues in collaboration with other compiler engineers. Triage and root-cause correctness and performance issues encountered while enabling ML models on EdgeTPU. Interact closely with model owners to influence their model architectures to make them run efficiently on the EdgeTPU.
Job description provided by Google.
View the original posting
