Lead Solutions Architect (United States)

Lead Architect, GPU Cluster Solutions | AI Infrastructure | Remote (US) | Up to $240k

A fast-growing AI infrastructure company deploying large-scale GPU clusters through a global partner network, moving fast from contract award into delivery, and scaling out the technical architecture team to match.

The Role

Designing here means starting from client requirements and NVIDIA Reference Architecture, then building out a real, deployable, supportable cluster: compute, storage, network, sparing, the lot.

Specifically:

• Designing GPU cluster configurations against NVIDIA Reference Architecture (HGX, NVL72) for every signed engagement
• Owning network topology and fabric selection, including InfiniBand, RoCE, and Ethernet, matched to each client's workload
• Building the hot/cold sparing plan that actually protects the SLA, not just satisfies a checkbox
• Folding in firewall, VPN, and dedicated circuit requirements into the design stage
• Adjusting designs around real-world site constraints, power, cooling, hardware lead times, working with Data Center Procurement and Supply Chain

Who they're looking for

• 7+ years in solutions architecture, network engineering, or systems engineering for GPU, HPC, or large-scale compute infrastructure
• Deep working knowledge of NVIDIA Reference Architecture and GPU cluster design principles
• Hands-on with InfiniBand, RoCE, and high-speed Ethernet fabric design, not just theory
• Experience building sparing strategies for infrastructure where downtime isn't an option
• Comfortable owning firewall, VPN, and dedicated circuit design as part of the network picture
• Bonus: high-speed shared storage experience, NVIDIA NCP familiarity, capacity or sparing modeling tools, prior design work for large enterprise or neocloud clients

Additional details

• Remote or hybrid, ideally US based, travel to data center partner sites required
• Up to $240k plus bonus and equity
• Reports to the VP, Deployments

If your background aligns or you know someone whose does, drop me a message!