Description

You will implement machine learning algorithms and models for large-scale deep learning training on NVIDIA supercomputers.

You will own the release pipeline.

Responsibilities

• Implement deep learning models tailored for NVIDIA supercomputers with a focus on high-performance networking.
• Profile, benchmark, and analyze models to identify optimizations for performance, efficiency, and accuracy.
• Design and implement scalable training pipelines and frameworks leveraging high-performance networking.
• Collaborate with software and hardware engineers to integrate networking solutions such as RoCE or InfiniBand.
• Analyze large-scale training results to resolve networking bottlenecks and optimize model outcomes.

Required Skills

• B.Sc. in Computer Science, Software Engineering, or 12+ years of equivalent experience.
• Deep learning expertise with practical experience in TensorFlow and PyTorch.
• Proficiency in CUDA programming for NVIDIA GPUs.
• Experience with networking libraries like NCCL and protocols including RoCE and RDMA.
• Strong understanding of parallel computing, distributed systems, and supercomputer architectures.
• Ability to profile and optimize deep learning workflows to resolve networking-related bottlenecks.
• Expertise in high-performance networking technologies like InfiniBand.

Preferred Skills

• Proven experience profiling and optimizing large-scale deep learning training on NVIDIA supercomputers.
• Expertise in optimizing networking parameters such as bandwidth, latency, or congestion control for deep learning workloads.

Key Skills

Education

Bachelor’s Degree