The ideal candidate will profile and analyze AI workloads on large GPUs and CPUs scale clusters for distributed Deep Learning LLM training focusing at the collectives communication and networking.
You will work and interact with many types of HW and platforms such as HCAs, Switches, CPUs, GPUs, and Systems.
You will experience with and develop performance analysis tools and methodologies to dive deeply into the details, understand performance expectation, limitations, and bottlenecks.
What you'll be doing:
Experience and research AI workloads and DL models specifically tailored for large-scale deep learning LLM training on NVIDIA supercomputers with a focus on High-performance networking.
Benchmarking, Profiling, and Analyzing the performance to find bottlenecks and identify areas of improvement and optimizations, with a strong emphasis on networking aspects.
Implement performance analysis tools.
Collaborating with many teams from HW to SW to provide performance analysis insights.
Define performance test planning , set performance expectations for new technologies and solutions, and work to reach the performance targets limits.
What we need to see:
B.Sc in Computer Science or Software Engineering.
5+ years of experience with high-performance Networking (RDMA, MPI).
Demonstrated Performance Analysis skills and methodologies.
Experience with our GPUs, CUDA library, deep learning frameworks like TensorFlow or PyTorch, combined with expertise in networking collective communication libraries (such as NCCL) and protocols (such as RoCE and RDMA).
Fast and self-learning capabilities with strong analytical and problem-solving skills.
Programming Languages: Python, Bash and C languages.
Experience with Linux OS distros.
Team player with good communication and interpersonal skills.
Ways to stand out from the crowd:
In-depth knowledge and experience with AI workloads and benchmarking for distributed LLM training.
Knowledge in CUDA, and NCCL libraries.
Knowledge in Congestion Control algorithms.
In-depth System knowledge and understanding (Intel / AMD / ARM CPUs, our GPUs, HCA, Memory, PCI).
Strong Performance Analysis skills and methodologies using modern tools.
