Senior HPC Network Architect | GPU Clusters

San Francisco Compute Company; San Francisco (CA); Hybrid; USD 200,000 - 280,000; Full time

Benefits offered by this job

Visa Sponsorships

401(k) matching

Medical, dental & vision

Job summary

San Francisco Compute is seeking an experienced Infrastructure Architect to design, deploy, and operate GPU clusters worldwide. You will lead networking strategy, implement zero-touch provisioning, and automate complex data-center deployments.

You will mentor engineers and work closely with customers to tailor solutions. The role requires 10+ years in GPU cluster networks, strong RoCEv2/InfiniBand knowledge, and a willingness to be on-site in San Francisco 3–4 days a week.

Qualifications

• 10+ years of hands-on management or architecture experience with at least one GPU cluster
• Deep understanding of Ethernet-based HPC networks (RoCEv2) and InfiniBand in CLOS/fat-tree topologies
• Experience building HPC network architectures (eBGP, fat-tree, VXLAN, MCLAG)
• Interest in zero-touch provisioning and automation for large multi-layer networks
• Strong documentation skills and ability to mentor junior engineers
• Willingness to work in SF office 3–4 days per week

Responsibilities

• Architect and deploy GPU clusters globally, ensuring reliability and scalability
• Participate in on-call rotations and troubleshoot network issues
• Drive automation to enable scalable deployments across data centers
• Mentor junior engineers and influence team culture and practices
• Collaborate with customers and internal teams to design custom regional cluster solutions

Skills

Networking architecture

RoCEv2 / InfiniBand

HPC networks

Automation

Documentation

Mentoring

SF office on-site

Tools

KVM

QEMU

libvirt

Kubernetes

Linux

Job description

San Francisco Compute is seeking an experienced Infrastructure Architect to design, deploy, and operate GPU clusters worldwide. You will lead networking strategy, implement zero-touch provisioning, and automate complex data-center deployments.

You will mentor engineers and work closely with customers to tailor solutions. The role requires 10+ years in GPU cluster networks, strong RoCEv2/InfiniBand knowledge, and a willingness to be on-site in San Francisco 3–4 days a week.