OpenAI
Software Engineer, Collective Communication
$380K–$500K
San Francisco · Full-time · Posted Apr 20, 2024
About the job
Designs and implements custom collective communication operations integrated with OpenAI’s training stack and supercomputer networks.
What you'll do
- Collaborate with ML researchers to design and implement efficient collective operations in C++ and CUDA.
- Ensure large training jobs use the different network transports effectively.
- Run simulations to inform future supercomputer network designs.
What they ask for
- Background in low-level, performance-critical software.
- Comfort writing performance-sensitive CPU and/or GPU code.
Nice to have
- Experience with collective communication.
- Prior distributed algorithms using RDMA or familiarity with network simulation techniques.
Pay
$380K – $500K USD base salary
Tools
C++, CUDA, RDMA
AI focus
Builds networking and collective communication infrastructure for large-scale model training.
Model-written summary · the original posting is authoritative