Senior System Software Engineer - GPU Performance
Core
Performance engineer optimizing communication libraries (NCCL, NVSHMEM, UCX) for Deep Learning and HPC on large-scale GPU clusters.
Role type
Senior System Software Engineer (GPU Performance)
Builds
High-performance communication libraries for AI and HPC applications
Domain
High Performance Computing, Artificial Intelligence, GPU Systems
Deliverable
production ML models
Required skills
Parallel programming, Performance benchmarking, Systems software fundamentals, C/C++ implementation, Debugging HW/SW stack, Python scripting, Container orchestration
Preferred skills
Infiniband/Ethernet networking (RDMA), CUDA programming, Deep Learning Frameworks (PyTorch, TensorFlow)
Technologies
NCCL, NVSHMEM, UCX, MPI, Infiniband, Ethernet, NVLink, PCIe, Kubernetes, SLURM, Ansible, Docker, CUDA, PyTorch, TensorFlow
Responsibilities
Conduct performance characterization on multi-GPU clusters, Study HW-SW interactions, Evaluate proof-of-concepts, Triage performance issues, Build performance analysis tools, Collaborate across time zones
Seniority
Senior, hands-on IC