Senior GPU & AI Infrastructure Architect

Referment; Greater London; On-site; GBP 90,000 - 120,000; Full time

Job summary

Referment is partnering with a cloud infrastructure provider to shape large-scale GPU platforms for customers across the UK. The role focuses on building enterprise GPU clusters, defining robust architectures and detailed BOMs.

You will translate workloads into Kubernetes-based or bare-metal solutions, lead discovery sessions, and drive proofs of concept with benchmarking to validate performance and cost considerations.

Qualifications

• Minimum five years in solution architecture, systems engineering or technical pre-sales for HPC/AI infra.
• Deep knowledge of NVIDIA HGX or DGX systems and GPU interconnects.
• Experience with InfiniBand/RoCE networks and GPU-direct technologies.
• Hands-on with Kubernetes for workload orchestration or bare-metal with Slurm/Ansible/Terraform.
• Proven ability to produce end-to-end designs and diagrams.

Responsibilities

• Design enterprise GPU clusters with high-level & low-level designs, rack and network diagrams, and detailed bills of materials.
• Shape scale-up and scale-out architectures across NVIDIA GPU platforms, NVLink and NVSwitch, high-speed InfiniBand and RoCE fabrics, storage, power and cooling.
• Translate customer workload, performance and commercial requirements into practical bare-metal, Slurm or Kubernetes-based solutions.
• Lead technical discovery sessions, architecture workshops and executive presentations, and contribute to complex proposals and RFP responses.
• Plan and oversee proofs of concept, using appropriate benchmarking to validate throughput, latency and distributed training performance.
• Work closely with enterprise customers, hardware vendors and internal engineering teams, feeding technical insight back into the platform roadmap.

Skills

Solution architecture

Systems engineering

Technical pre Sales

NVIDIA HGX/DGX knowledge

GPU interconnects

InfiniBand/RoCE networking

Kubernetes

Slurm

Ansible

Terraform

HLD/LLD design

Executive communication

Education

Degree in a relevant field

Tools

NVIDIA HGX/DGX systems

InfiniBand switches

RoCE fabrics

Job description

Referment is partnering with a cloud infrastructure provider to shape large-scale GPU platforms for customers across the UK. The role focuses on building enterprise GPU clusters, defining robust architectures and detailed BOMs.

You will translate workloads into Kubernetes-based or bare-metal solutions, lead discovery sessions, and drive proofs of concept with benchmarking to validate performance and cost considerations.