About this role
NVIDIA’s AI Networking Codesign and Benchmarking R&D group seeks a Principal Developer to analyze and optimize AI workloads on large-scale GPU and CPU clusters used for distributed deep learning and LLM training and inference. The role focuses on collective communication and high-performance networking across hardware systems, machine learning frameworks, and communication and computing libraries.
Responsibilities
- Characterize AI workloads and deep learning models on NVIDIA supercomputers, with particular attention to distributed systems and networking.
- Benchmark and profile performance, identify bottlenecks, and recommend optimizations. Develop a PyTorch trace-based toolset for profiling, analysis, replay, benchmarking, debugging, and network-system co-design for LLM workloads.
- Collaborate with hardware and software teams to share performance insights. Define performance test plans and expectations for new technologies and work toward performance targets.
Required qualifications
- B.Sc. in Computer Science or Software Engineering, or equivalent experience, and 15+ years of experience with high-performance networking, including RDMA, MPI, NCCL, or SHARP.
- Demonstrated performance-evaluation ability; experience with NVIDIA GPUs and CUDA; knowledge of TensorFlow or PyTorch; and expertise in collective communication libraries such as NCCL and protocols such as RoCE and RDMA.
- Proficiency in Python, Bash, and C++; experience in container-based development; strong analytical, problem-solving, learning, communication, and teamwork skills.
Preferred qualifications
- Hands-on experience benchmarking AI workloads for distributed LLM training; knowledge of PyTorch, CUDA, and NCCL; broad systems knowledge covering Intel, AMD, or ARM CPUs, NVIDIA GPUs, HCAs, memory, and PCI; and strong performance-evaluation skills using contemporary tools.
Locations listed are Santa Clara, California, and remote roles in Texas, Colorado, and Washington. The listed base salary ranges are 272,000–431,250 USD for Level 6 and 320,000–488,750 USD for Level 7, with equity and benefits eligibility. Applications will be accepted at least until June 16, 2026.
Skills for this role
High-performance networkingDistributed systemsLLM trainingLLM inferenceDeep learningPerformance benchmarkingPerformance profilingCollective communicationRDMAMPINCCLSHARPRoCECUDAPyTorchTensorFlowPythonBashC++Container-based developmentNVIDIA GPUsPerformance analysisAnalytical problem-solvingCommunicationCollaboration