全职 AI Infrastructure Architect – Hardware - Physical-Layer Networking 工作, 薪水, Gateway Search 公司招聘中 - Ricebowl

AI Infrastructure Architect – Hardware & Physical-Layer Networking

Location: Singapore | Full-time

About Our Client

We are partnering with a fast-growing AI infrastructure provider that builds and operates high-performance GPU cloud and AI-ready data centers for AI labs and enterprises. As they scale their capacity, they are hiring a hands-on architect to lead the hardware design of their next generation of GPU clusters.

The Opportunity

You will own the hardware architecture of large-scale AI data centers end to end, from cluster topology down to the last optical transceiver. This role suits someone who wants real architectural ownership and still enjoys being close to the physical build.

What You’ll Do

• Own end-to-end hardware architecture for AI data centers, covering compute, GPU and accelerators, networking, storage, racks and supporting infrastructure, from selection through deployment.
• Design GPU cluster topology across intra-node interconnect (NVLink/NVSwitch), inter-node fabric (InfiniBand/RoCE) and non-blocking architectures (Fat-Tree / Spine-Leaf).
• Own the physical layer:
• – switch selection
• – port and uplink planning
• – optical transceiver selection and compatibility
• – DAC/AOC vs. fiber strategy
• – rack- and row-level cable management
• Evaluate, benchmark and validate servers and key components (CPU, GPU, memory, NICs, storage, PSUs), and produce detailed BOMs covering switches, optics and cabling.
• Oversee hardware delivery: rack-and-stack, structured cabling, power-on and acceptance testing.
• Lead physical-layer troubleshooting, including link errors, optical signal issues and cabling faults.
• Track emerging technologies such as next-gen GPUs and accelerators, co-packaged optics (CPO) and liquid cooling, and evolve the architecture for performance and cost efficiency.

What You’ll Bring

• Bachelor’s degree or above in Computer Science, Electrical Engineering, Communications, Automation or a related field.
• 5+ years in server or data center hardware architecture, or in infrastructure solutions.
• Deep understanding of server architecture (CPU, GPU, memory, PCIe, NVLink) and mainstream GPU platforms, such as 8-GPU servers and NVIDIA HGX.
• Strong hands-on expertise in data center network hardware:
• – InfiniBand/RoCE
• – switch selection and configuration
• – optical transceivers (compatibility, reach, form factor)
• – DAC/AOC vs. fiber trade-offs
• – non-blocking topology design
• Solid grasp of data center infrastructure: power distribution, cooling (especially liquid cooling), rack layout and high power-density deployments.
• Proven ability to independently drive hardware selection, performance testing and capacity planning, and to deliver complete architecture proposals and BOMs.
• Strong vendor management, cross-functional collaboration and technical documentation skills.

Strongly Preferred

• Experience designing, delivering or operating large-scale GPU clusters of hundreds to thousands of GPUs.

Why This Role

• Architect-level ownership at a company at the center of the AI compute build-out.
• Hands-on exposure to the latest GPU platforms, high-speed fabrics and liquid-cooled, high-density deployments.
• High-visibility role with direct influence on infrastructure strategy and cost efficiency.

All applications will be treated with strictest confidence. We regret to inform you that only shortlisted candidates will be notified.

EA Licence No: 19C9807

Registration No: R22110556 (ZHOU DONGYANG)