OpenAI
Software Engineer, Workload Enablement
$293K–$385K
San Francisco +1 · Hybrid · Full-time · Posted Mar 28, 2026
- Location
- San Francisco
Seattle - Experience
- 5+ years in one or more of ML systems, performance engineering, distributed systems, or HPC.
- Education
- BS in CS/EE (or equivalent practical experience).
About the job
Enables production AI workloads on new platforms by porting and validating workloads, building end-to-end benchmarks, and diagnosing system performance and reliability.
What you'll do
- Port and validate inference and training workloads on new platforms, ensuring correctness, performance, and stability.
- Build repeatable benchmarks and test harnesses for end-to-end workload behavior, regression detection, and platform stress testing.
- Analyze distributed workload performance and collaborate with engineering teams and vendors to resolve platform issues.
What they ask for
- BS in CS/EE or equivalent practical experience.
- 5+ years in one or more of ML systems, performance engineering, distributed systems, or HPC.
- Hands-on experience with PyTorch and modern LLM training/inference stacks, distributed training concepts, RDMA, and NCCL or RCCL.
- Proficiency in Python and strong profiling/debugging skills.
Nice to have
- Experience building workload-shaped benchmarks and stress/fault tests that correlate with production behavior.
- Experience validating workloads in Kubernetes or hands-on lab experience with early hardware.
Pay
$293K – $385K USD base salary
Tools
PyTorch, NCCL, RCCL, Python, C++, CUDA, HIP, Kubernetes
AI focus
Ports and validates AI inference and training workloads, including modern LLM workloads, on new platforms; benchmarks and analyzes their end-to-end performance.
Model-written summary · the original posting is authoritative