Lead GPU Cluster Solution Architect

Axe Compute; Miami (FL); On-site; USD 140,000 - 170,000; Full time

Job summary

Axe Compute in Miami is seeking a Lead Architect, GPU Cluster Solutions to craft buildable GPU cluster designs aligned with NVIDIA Reference Architecture for signed engagements. You will translate client requirements into a complete design covering compute, storage, and networking.

Lead the sparing, availability, and site-constraint adaptations, collaborating with deployments leadership to ensure the design translates into actionable project plans and validated acceptance criteria.

Qualifications

• 7+ years in solutions architecture or systems engineering for GPU/HPC infrastructure.
• Deep knowledge of NVIDIA Reference Architecture (HGX, NVL72).
• Hands-on experience with InfiniBand, RoCE, and high-speed Ethernet.
• Experience incorporating firewall, VPN, and dedicated circuit requirements into designs.
• Experience with high-speed shared storage solutions.

Responsibilities

• Design GPU cluster configurations against NVIDIA Reference Architecture for each signed engagement.
• Translate client requirements into a complete bill of design with compute, storage, and networking.
• Design network topology and fabric selection (InfiniBand/RoCE/Ethernet) per workload.
• Incorporate internet/VPN/firewall connectivity into cluster designs.
• Formulate hot/cold sparing and availability plans for SLAs.
• Adapt designs to site power, cooling and space constraints with procurement.
• Collaborate with VP Deployments to translate designs into buildable project plans.
• Support acceptance test design and criteria for as-designed architecture.

Skills

GPU HPC infrastructure

Solutions architecture

Network engineering

Data center design

Cross-functional collaboration

Tools

InfiniBand

RoCE

Ethernet fabric design

Firewall/VPN/dedicated circuit

NVIDIA Reference Architecture (HGX/NVL72)

Job description

Axe Compute is seeking a Lead Architect, GPU Cluster Solutions to design GPU cluster configurations for prospective and signed engagements, translating client requirements and NVIDIA Reference Architecture into a buildable, supportable design, spanning compute, storage, networking, software and spares strategy for support.

ROLE AT A GLANCE

• Mandate: Own end-to-end technical design of GPU cluster deployments, from client requirements through NVIDIA Reference Architecture compliance and sparing strategy.
• Scope: Cluster design, network architecture (InfiniBand/RoCE/Ethernet), sparing and spares planning, connectivity design (internet/VPN/firewall, dedicated circuits), design adjustments for site and hardware constraints.
• Key Outcomes: Designs that meet client SLAs and NVIDIA Reference Architecture standards, sparing plans that protect uptime commitments, designs that account for real-world site and hardware lead-time constraints.

WHAT YOU WILL OWN

Cluster Design & Reference Architecture

• Design GPU cluster configurations (compute, storage, networking) against NVIDIA Reference Architecture for each signed engagement.
• Translate client technical requirements into a complete bill of design, including all necessary compute, storage, and networking components.
• Design network topology and fabric selection, including InfiniBand, RoCE, and Ethernet options, appropriate to each client's workload and performance requirements.
• Incorporate internet, VPN, and firewall connectivity requirements into cluster designs.
• Design dedicated point-to-point network requirements where needed, including protected optical circuits and similar dedicated connectivity.

Sparing & Availability Strategy

• Formulate and own the hot/cold sparing plan for each deployment to meet contracted SLA commitments.
• Adjust sparing and design assumptions based on data center power/cooling parameters and hardware lead-time constraints.

Design Adaptation & Site Constraints

• Adjust cluster designs to fit site-specific power, cooling, and space constraints identified by the Data Center Procurement and Operations Director.
• Work with Supply Chain to align design decisions with realistic hardware delivery timing.

Cross-Functional Collaboration

• Partner with the VP, Deployments and Deployment Program Manager to ensure designs translate cleanly into buildable, trackable project plans.
• Support acceptance test design and criteria definition, ensuring test procedures validate the as-designed architecture.

REQUIRED QUALIFICATIONS

• 7+ years in solutions architecture, network engineering, or systems engineering supporting GPU, HPC, or large-scale compute infrastructure.
• Deep working knowledge of NVIDIA Reference Architecture (HGX, NVL72) and GPU cluster design principles.
• Hands-on experience with InfiniBand, RoCE, and high-speed Ethernet fabric design.
• Experience incorporating firewall, VPN, and dedicated circuit (e.g., protected optical) requirements into network designs.
• Experience with high-speed shared storage solutions (e.g., Weka, Vast, DDN).

PREFERRED QUALIFICATIONS

• Experience designing clusters for large enterprise clients or neoclouds, not just internal infrastructure.
• Familiarity with NVIDIA NCP program requirements and certification processes.
• Experience with capacity or sparing modeling tools.