Senior Infrastructure Architect

• We will have you own network architecture across GPU fabric, InfiniBand, RoCE v2, Ethernet leaf-spine, edge connectivity, peering, observability, deployment standards, and operational handover.
• We will have you help define the standards future engineers will build and operate against.
• We will have you design and support network fabrics for large-scale GPU compute infrastructure.
• We will have you address network issues that can impact customer training workloads, including routing, congestion, and fabric design challenges.
• We will have you set the foundation for a new senior network function in our organization.

Technologies:

• AI
• Cloud
• Cisco
• Ethernet
• Fabric
• Hardware
• InfiniBand
• Support
• Network
• Architect

More:

We are a stealth AI infrastructure company building large-scale GPU compute infrastructure for frontier AI customers. We are creating AI factories at industrial scale, where network fabric is as critical as compute itself. This is our first senior network hire, and the role will shape the standards and foundation for future engineering. If you have built, deployed, or operated network fabrics for serious GPU infrastructure, this is a rare opportunity to help define one of the most ambitious AI infrastructure builds in the market.

• We require deep experience in GPU cluster or HPC deployments.
• We require strong production experience with InfiniBand.
• We require experience operating RoCE v2 at scale.
• We require experience with Ethernet fabrics, including BGP, ECMP, and low-latency operations.
• We require IPv6 and public ASN experience.
• We require multi-vendor networking experience across environments such as NVIDIA Mellanox, Arista, Juniper, Cisco, whitebox, or ODM hardware.
• We require experience from a neo-cloud, hyperscaler, major vendor, GPU infrastructure company, HPC platform, or AI infrastructure provider.
• We require direct experience working on real GPU deployments.
• We require a background in relevant AI infrastructure, hyperscale, vendor, or HPC environments.
• We are not seeking single-vendor specialists or traditional enterprise networking profiles.