Senior Network Architect — AI GPU Infra (InfiniBand)

GMI Cloud, Inc; United States; On-site; USD 180,000 - 240,000; Full time

Job summary

GMI Cloud is seeking an experienced Network Architect to join the Global Infrastructure team. You will lead the design, deployment, and day-to-day operations of high-performance network infrastructure supporting GPU-accelerated AI/ML workloads.

Preferred Location: US, Taipei, APAC. Responsibilities include designing InfiniBand/RoCE fabrics, planning global data-center networks, and driving automation with Ansible and REST APIs. Travel to data centers will be required.

Qualifications

• Bachelor's degree in CS or related field.
• 10+ years networking experience with 3+ years in InfiniBand/RoCE fabrics for large GPU clusters.
• Hands-on InfiniBand HDR/XDR fabrics, subnet manager, routing, and P_Key knowledge.
• Experience with RDMA, RoCE, and Ethernet lossless frameworks (PFC/ECN/DCB).
• Strong troubleshooting: packet analysis, counters, link diagnostics, congestion root-cause.
• Familiarity with NCCL, MPI, GPUDirect RDMA and GPU networking patterns.
• Experience with NVMe-oF, NFS over RDMA, GPUDirect Storage.
• Hands-on with Mellanox/NVIDIA Spectrum, Cisco hardware.
• TCP/IP, VLANs, L3 routing, MTU/Jumbo frames, L2/L3 troubleshooting.
• Monitoring tools (Prometheus, Grafana, SNMP, ELK) and network automation (Ansible, REST APIs).
• Security knowledge: DDOS, IDS; cross-functional collaboration; bilingual in English/Chinese preferred.

Responsibilities

• Architect, design, and implement high-performance network fabrics for GPU clusters.
• Plan and deploy global data center networks, WAN, core, DC, firewalls and load balancers.
• Build networks for Compute, Storage, Inband, Management, and OOB using InfiniBand and RoCEv2.
• Lead hardware lifecycle: HBAs, NICs, RDMA NICs, switches and firmware.
• Configure RDMA, QoS, ECN, PFC and Flow Control to optimize MPI/NCCL.
• Operate day-to-day: provisioning, patches, upgrades, monitoring.
• Troubleshoot production incidents, drive RCA, implement fixes.
• Develop automation, runbooks, and capacity planning (Ansible, REST APIs).
• Collaborate with compute, storage, platform, and SRE teams; validate end-to-end perf.
• Identify providers and solutions; enforce network security and segmentation.
• Maintain diagrams, cabling maps, baselines, and procedures; mentor juniors.
• Occasional travel to GMI data centers.

Skills

InfiniBand/RoCE design

RDMA

Ethernet lossless frameworks

NCCL/MPI

GPU-communication patterns

Network troubleshooting

Automation scripting

Ansible

REST APIs

Prometheus

Grafana

SNMP

ELK

Mellanox/NVIDIA Spectrum

Cisco switches/routers

TCP/IP

BGP/OSPF

GPUDirect RDMA

Network security

Education

Bachelor’s degree in Computer Science or related field

Tools

Ansible

Prometheus

Grafana

SNMP

ELK

Mellanox/NVIDIA Spectrum

Cisco

Job description

GMI Cloud is seeking an experienced Network Architect to join the Global Infrastructure team. You will lead the design, deployment, and day-to-day operations of high-performance network infrastructure supporting GPU-accelerated AI/ML workloads.

Preferred Location: US, Taipei, APAC. Responsibilities include designing InfiniBand/RoCE fabrics, planning global data-center networks, and driving automation with Ansible and REST APIs. Travel to data centers will be required.