Description
The role designs and implements collective communication algorithms and a topology-aware transport layer for Intel’s AI accelerator fabric, optimizing performance across multi-node pods and ensuring correctness and numerical determinism. It requires 5+ years of AI/Systems/HPC software experience in C++, strong concurrency and lock-free design, collective-library expertise, and performance-engineering skills. The position is an experienced hire based in Israel, with work locations in Haifa and Petah-Tikva, and offers a hybrid work model.
