Pragmatike

GPU Server Hardware Validation Engineer

China / Guangzhou. Posted 2 months ago

The role

This person will own quality control across manufacturing sites and run the hardware test lab. They will validate GPU compute and interconnect performance and diagnose hardware, firmware, and driver-level issues.

What you will do

  • Own quality control across manufacturing sites on the factory floor
  • Run the hardware test lab for board and system bring-up, validation and stress testing
  • Oversee assembly and end-to-end testing of GPU servers
  • Define and maintain test procedures, acceptance criteria and QC checkpoints
  • Diagnose hardware, firmware and driver-level issues
  • Drive root-cause analysis with design and manufacturing teams
  • Validate GPU compute and interconnect performance under Linux
  • Document failures, yields and test results
  • Coordinate with manufacturing partners and local suppliers in English and Mandarin

Required qualifications

  • Hands-on experience testing and validating GPU/HPC servers
  • Strong Linux skills at the system and hardware level
  • Firmware knowledge: BIOS/UEFI, BMC/IPMI, GPU and NIC firmware, flashing and troubleshooting
  • Working knowledge of the NVIDIA GPU stack: drivers, CUDA, NCCL
  • Understanding of server hardware: motherboards, PCIe topology, power, cooling, storage and networking components
  • Experience with QC or manufacturing test processes in a hardware production environment
  • Fluent English
  • Fluent Mandarin
  • Willingness to work on-site in Hong Kong, on the factory floor and in the lab

Nice to have

  • Experience with containerized or micro-datacenter deployments
  • Familiarity with high-speed interconnects (InfiniBand, RoCE, NVLink)
  • Scripting skills (Bash/Python)
  • Exposure to hardware supply chains or contract manufacturers in Hong Kong / Southern China
  • Cantonese
  • Startup or hyper-growth experience

About the company

Pragmatike is recruiting on behalf of a European deep-tech company building a sovereign, energy-efficient alternative to traditional cloud providers. The client designs and manufactures its own HPC infrastructure end to end.