NVIDIA
Software Engineering Manager - GPU Communications Libraries
$184K–$356K
US, CA, Santa Clara · Manager · Full-time · Posted Oct 7, 2026
- Experience
- 10+ years overall in the software industry and 4+ years of management experience.
- Education
- BS, MS, or Ph.D. in CS, CE, EE (related technical field) or equivalent experience.
About the job
Leads the team developing and managing NVSHMEM and UCX libraries, including technical design and implementation, project execution, library quality and performance, and partner collaboration.
What you'll do
- Lead, mentor, and grow the engineering team; plan and execute projects and oversee library quality and performance.
- Participate in feature design and implementation and collaborate with partners and internal teams on requirements and product roadmaps.
- Identify improvements to processes, infrastructure, and practices.
What they ask for
- 10+ years of software-industry experience specializing in HPC networking or system software.
- 4+ years of management experience.
- Prior systems software, communication runtime, or high-performance networking software development experience, including taking complex features or products through the full product life cycle.
- Strong computer architecture, operating systems, HW-SW interaction, and performance analysis knowledge; excellent C/C++ programming and debugging skills in Linux.
Nice to have
- Experience with parallel programming models, communication runtimes, and CUDA, MPI, OpenMP, OpenACC, or pthreads.
- Background in RDMA, high-performance networking, network architecture/topologies, HPC and ML/DL, or Deep Learning frameworks such as PyTorch and TensorFlow.
Pay
184,000 USD - 287,500 USD (base salary, Level 2); 224,000 USD - 356,500 USD (base salary, Level 3)
Tools
NVSHMEM, UCX, C/C++, Linux, MPI, NCCL, CUDA, PyTorch
AI focus
The libraries support deep learning and HPC applications; NVIDIA also states that it uses AI tools in its recruiting processes.
Model-written summary · the original posting is authoritative