Senior Networking Engineer - HPE/AI Networking

ARC; San Jose (CA); On-site; USD 150,000 - 240,000; Full time

Job summary

ARC in San Jose designs, deploys, and operates high-performance fabrics underpinning large NVIDIA GPU clusters, including InfiniBand, RoCE, RDMA, and 400G/800G networking.

The candidate will architect scalable leaf/spine topologies, deploy cutting-edge networking infra, and troubleshoot RDMA, NCCL, latency, and congestion while driving automation with Python or Ansible. Prior data-center AI infrastructure experience is required.

Qualifications

• 5+ years in large-scale data center or AI infrastructure networking.
• Hands-on InfiniBand Mellanox and RoCE networking experience.
• Strong RDMA, RoCE, and NCCL performance validation skills.
• Experience with 100G/200G/400G/800G networking.
• Good Ethernet, TCP/IP, BGP, VLAN, DC networking knowledge.
• Automation with Python or Ansible.

Responsibilities

• Design and operate high-performance network fabrics for large GPU clusters.
• Architect InfiniBand and RoCE networks for distributed AI training.
• Design scalable leaf/spine GPU cluster topologies.
• Deploy and operate 400G/800G networking infrastructure.
• Work with NVIDIA Mellanox ConnectX NICs and NVIDIA networking platform.
• Troubleshoot RDMA, NCCL, latency, congestion, and fabric performance.
• Optimize GPU-to-GPU communication and distributed training performance.
• Develop network provisioning, automation, monitoring, and remediation.
• Analyze bottlenecks across GPU, storage, and cluster infrastructure.
• Manage network hardware lifecycle: switches, optics, cabling, sparing.
• Maintain network docs and change records.

Skills

Network engineering

RDMA/NCCL

InfiniBand

RoCE

Python automation

Ansible

Network troubleshooting

Tools

Python

Ansible

Job description

Design, deploy, and operate the high-performance fabrics behind large-scale clusters — InfiniBand, RoCE, RDMA, 400G/800G networking, NVIDIA networking technologies, and distributed AI workloads.

Key Responsibilities

• Design and operate high-performance network fabrics for large-scale NVIDIA GPU clusters.
• Architect InfiniBand and/or RoCE networks for distributed AI training.
• Design scalable leaf/spine and GPU cluster network topologies.
• Deploy and operate 400G/800G networking infrastructure.
• Work with NVIDIA/Mellanox ConnectX NICs and NVIDIA networking platform.
• Troubleshoot RDMA, NCCL, packet loss, latency, congestion, and fabric performance issues.
• Optimize GPU-to-GPU communication and distributed training performance.
• Build network provisioning and configuration automation.Develop network monitoring, telemetry, alerting, and automated remediation.
• Analyze network bottlenecks across GPU, storage, and cluster infrastructure.
• Manage network hardware lifecycle: switches, optics, cabling standards and sparing
• Maintain network documentation, diagrams, and change records

Required Qualifications

• 5+ years of network engineering experience in large-scale data center, cloud, HPC or AI infrastructure.
• Deep understanding and hands-on experience with InfiniBand Mellanox and RoCE networking technologies.
• Strong knowledge of RDMA/RoCE and NCCL-based performance validation
• Experience with 100G/200G/400G/800G networking
• Strong understanding of Ethernet, TCP/IP, BGP, VLAN, routing, switching, and data center networking.
• Strong troubleshooting and network performance analysis skills.
• Experience automating infrastructure using Python, Ansible, or similar tools

Preferred Qualifications

• Spectrum-X / SONiC experience
• Strong Ethernet fundamentals: BGP, OSPF, EVPN/VXLAN, L2/L3 design
• Experience supporting large AI training clusters (1,000+ GPUs)
• Relevant certifications (CCNP/JNCIP or equivalent)