jobs

Software Engineer, Collective Communication

About the TeamThe Workload Networking team is responsible for the collective communication stack used in our largest training jobs. Using a combination of C++ and CUDA we work on novel collective communication techniques that enable efficie

Company

OpenAI

Location

San Francisco

Status

Open

Posted

2024-04-20T03:52:24.444+00:00

OpenAI's Workload Networking team is hiring a software engineer to build the collective communication stack for its largest training jobs. The work involves designing and implementing custom networking collectives in C++ and CUDA, keeping training runs efficient across the network transports in OpenAI's supercomputers, and running simulations to inform future network designs. Candidates should have written low-level, performance-critical CPU or GPU code; experience with RDMA-based distributed algorithms, network simulation, or collective communication is helpful. Based in San Francisco, three days a week in the office.