Staff Datacenter Networking for GPU AI Infrastructure

Prime Intellect AI; San Francisco (CA); On-site; USD 150,000 - 300,000; Full time

Job summary

Prime Intellect AI is building an open frontier AI stack and a unified platform for large-scale GPU training, inference, and tooling. You will design and operate networks that connect massive GPU clusters, ensuring reliability and scalable performance across training fabrics, storage networks, and management connectivity.

You will collaborate with hardware and ML teams to set standards, automate provisioning, and monitor fabric health.

Qualifications

• 3+ years of production datacenter networking experience.
• Strong understanding of Ethernet, TCP/IP, routing, switching, and redundant network design.
• Hands-on experience with high-performance GPU networking using InfiniBand or RoCE.
• Experience troubleshooting network problems across Linux hosts, NICs, switches, and links.
• Ability to automate network operations with Python, Ansible, or comparable tools.

Responsibilities

• Design and deploy scalable datacenter network topologies for GPU training, inference, storage, and management traffic.
• Configure and operate high-performance Ethernet/RoCE and InfiniBand fabrics with clear standards for routing, redundancy, and capacity.
• Automate network provisioning, configuration validation, upgrades, and rollback procedures.
• Diagnose packet loss, congestion, link failures, and collective communication performance across hosts and switches.
• Benchmark end-to-end network performance with infrastructure and ML teams, translating workload needs into measurable acceptance criteria.
• Build monitoring for port health, errors, utilization, congestion, and fabric topology; improve incident response and runbooks.
• Partner with datacenter operators and hardware vendors on cabling, optics, deployment readiness, and failure resolution.

Skills

Datacenter networking

InfiniBand/RoCE

TCP/IP routing

Linux networking

Python/Ansible automation

Job description

Prime Intellect AI is building an open frontier AI stack and a unified platform for large-scale GPU training, inference, and tooling. You will design and operate networks that connect massive GPU clusters, ensuring reliability and scalable performance across training fabrics, storage networks, and management connectivity.

You will collaborate with hardware and ML teams to set standards, automate provisioning, and monitor fabric health.