Software Engineer, Collective Communication and Network Performance
Core
Develop software tools to automatically detect and diagnose collective communication and network performance problems in AI workloads across thousands of GPUs.
Role type
Senior IC software engineer (collective communication & network performance)
Builds
Production software for NCCL and network fabric diagnostics
Domain
AI systems, High Performance Computing (HPC), GPU clusters
Deliverable
production ML models
Required skills
C++, Python, NCCL, RDMA (InfiniBand, RoCE), network topology analysis, debugging multi-node systems, telemetry ingestion, regression testing
Preferred skills
Contributions to NCCL/UCX/libfabric/NVSHMEM, experience with NIXL/DeepEP/NVSHMEM, EFA fabrics, production root-cause analysis, debugging training/inference jobs >1000 GPUs
Technologies
NCCL, RDMA, InfiniBand, RoCE, NVLink, C++, Python
Responsibilities
Develop detection algorithms for communication performance issues; build analysis infrastructure to ingest and correlate per-collective records; define data requirements for profiler/tuner plugins; reproduce and verify communication problems; extend coverage to point-to-point paths
Seniority
Senior, hands-on IC