Senior GPU Infrastructure Engineer — InfiniBand / Linux

Description

The company is building GPU infrastructure in the U.S. for AI training and inference, with planned deployments of NVIDIA B300 and RTX PRO 6000 servers.

The team needs a senior engineer who works hands-on to design, deploy, validate and run GPU clusters. You will work directly with the founding team, server manufacturers, networking providers and data-center partners.

Responsibilities

• Check server and networking bills of materials, network topology, cabling plans and hardware compatibility before each deployment.
• Deploy and troubleshoot NVIDIA/Mellanox InfiniBand and Ethernet networks used for management, storage and customer connectivity.
• Set up Linux servers, GPU drivers, NIC firmware and supporting infrastructure within the agreed customer service scope.
• Confirm GPU health and cluster performance with tools such as DCGM, NCCL tests and InfiniBand diagnostics.
• Set acceptance criteria, record test results and support the customer handover.
• Build monitoring, alerting, configuration backups and operational runbooks.
• Work with OEM support and data-center remote-hands teams to resolve hardware, network and performance faults.
• Take part in planned deployments, maintenance and the agreed on-call support.

Requirements

• You have 5+ years of relevant experience in infrastructure, Linux systems, networking or data-center engineering.
• You have hands-on experience deploying or running multi-node GPU clusters.
• You have practical InfiniBand experience, covering fabric configuration, subnet management, link diagnostics and performance troubleshooting.
• You are strong in Linux administration and Ethernet networking fundamentals.
• You can tell apart network, driver, hardware and workload causes of performance problems.
• You script or automate with Python, Bash or Ansible.
• You write clear documentation, communicate well and own technical delivery from start to finish.
• Experience with H100, H200, A100 or comparable GPU clusters is welcome; prior B300 experience is not required.
• Preferred: NVIDIA NCP-AIN or NCP-AII certification.
• Preferred: Cisco CCNP Data Center or CCIE Data Center certification.
• Preferred: experience with NVIDIA UFM, ConnectX adapters, RDMA and GPU cluster monitoring.
• Preferred: experience with bare-metal provisioning, customer acceptance testing or infrastructure operations at a GPU cloud provider.
• Certifications are a plus, but demonstrated deployment and troubleshooting experience is the primary selection criterion.

Conditions

You will be part of an agreed on-call support arrangement.