Senior Networking Performances Architect

Our technology has no boundaries! NVIDIA is building the world’s most groundbreaking and state-of-the-art accelerated compute platforms for the world to use. It’s because of our work that scientists, researchers, and engineers can advance their ideas.

We are seeking a highly motivated architect to join our team of experts and help shape the future of high-performance and ML/AI computing. Our next-generation NVL systems will be at the forefront of connecting and powering the world's most advanced compute clusters, from supercomputers used in AI research to high-performance clusters used at almost every industry today, such as car and pharmaceutical.

Responsibilities:

• Analyse the performances of novel network features and their impact on running AI workloads on the next large-scale, high-performance computing (HPC) clusters.
• Develop network behaviour models for simulations and performance analysis and prioritise future networking features development.
• Generate insights that focus on new generations of products to remain state of the art for AI.
• Collaborate with cross-functional teams, including deep learning, simulation, architecture, and research.
• Become an expert in AI networking on high-performance computing.

Requirements:

• B.Sc., M.Sc., or Ph.D. degree in Computer Science, Computer Engineer, or Electrical Engineer.
• 5+ years of system engineering, system architecture, or performance research experience.
• Deep understanding of complex system behaviour.
• Possess strong problem-solving and critical thinking skills.
• Ability to work and operate in a highly dynamic environment, be a team player, and collaborate with multiple groups in the organisation.
• Good knowledge in Python.

Nice to Have:

• Experience in development of simulation environments.
• Understanding of large-scale networks behaviour and the effect of distributed computing workloads on the network.
• Understanding of network protocols (InfiniBand, IP, TCP, RoCE).
• Good knowledge of AI models.