The role involves architecting, implementing, and maintaining a low-level networking library designed to power distributed inference and training. The contributor will design scalable communication systems to coordinate thousands of processors efficiently and reliably. Responsibilities include optimizing protocols, synchronization strategies, and data movement to maximize hardware performance. The individual will integrate networking APIs into the broader programming model while collaborating with hardware and software teams. Additionally, the role requires helping to define the long-term architecture of the distributed systems stack.

Candidates must be strong systems engineers with deep experience in C or C++. Proficiency in low-level or bare-metal environments is required. Applicants should possess a passion for hardware-software interaction, performance tuning, and protocol-level optimization. The role requires the ability to reason from first principles and challenge industry conventions. Employment is contingent upon the applicant's eligibility to access U.S. export-controlled technology.