Senior Solutions Architect – AI Infrastructure Networking

Description

The team is looking for a senior solutions architect with deep networking expertise to join the Office of the CTO's Voyager group. You will take ownership of the network design for k0rdent AI, the company's platform for building and running GPU clouds and AI factories, covering everything from bare metal to models.

You will decide how large GPU clusters are interconnected, how tenants are kept apart, and how workloads reach the network from containers, virtual machines and bare metal nodes. Those designs become published reference architectures, code and proofs of concept tested on real hardware with customers and partners.

Research is central to this role. You will look at designs and technologies before customers request them, build prototypes that may never ship, and question accepted practice when something better exists. Your work will influence the future direction of the product, not just current deployments.

The job mixes research with hands-on, external work. You will divide your time between exploring new ideas, writing architecture, validating it in the lab, and explaining it to engineers, executives and conference audiences. You will collaborate with Solutions Architects, Partner Management and Engineering across many countries and time zones.

You will work with customers and partners who run GPU cloud, neocloud, sovereign cloud and AI-first enterprise environments. The company is a Kubernetes-native AI infrastructure provider, part of IREN, that helps organisations run scalable, secure infrastructure on-premises, in the cloud, at the edge and in sovereign data centers, with a focus on open source and avoiding lock-in.

Responsibilities

• Design and publish network reference architectures and solution designs for k0rdent AI, from a single rack up to clusters of several thousand GPUs.
• Define east-west compute fabrics on InfiniBand and on RoCEv2 Ethernet, including rail-optimised designs, fat-tree and Clos layouts, oversubscription ratios and failure domains.
• Design front-end, storage and management networks, and how they link to customer data centers, public clouds and hybrid setups.
• Specify tenant isolation across the stack using InfiniBand partitions (PKeys), VRFs and EVPN-VXLAN on Ethernet, and Kubernetes network policy.
• Document scale limits and trade-offs covering cost, performance, operability and vendor lock-in, with reasoning customers can follow.
• Define how Linux hosts expose NICs, DPUs and SuperNICs to workloads, covering PCI passthrough, SR-IOV virtual functions, IOMMU groups, NUMA layout and GPU-NIC affinity.
• Design Kubernetes networking for AI workloads, including primary and secondary CNIs, Multus, SR-IOV and RDMA device plugins, the NVIDIA Network Operator and Dynamic Resource Allocation (DRA).
• Design virtual machine networking on KubeVirt, including passthrough and SR-IOV for GPU and RDMA traffic.
• Track and assess emerging AI networking technologies and standards, such as Ultra Ethernet (UEC), SuperNIC and DPU offloads, scale-up and scale-across fabrics, and Kubernetes DRA network drivers.
• Prototype alternative designs in the lab, including ones no customer has asked for, and benchmark them against current practice.
• Challenge established or customer-preferred designs when the evidence supports a better option, and argue the case with data.
• Publish research notes and design proposals internally, and external papers, blog posts or talks where suitable.
• Build and run proofs of concept with customers and partners in the company lab hardware and on customer sites, and report results against agreed success criteria.
• Write automation and tooling in Python, Go, Bash, Ansible, Helm, Kubernetes manifests or Terraform so the reference architectures can be reproduced.
• Run validation and benchmarks on fabrics and host configurations, such as NCCL tests, perftest and ibwritebw, and turn the results into design guidance.
• Act as the network subject matter expert in customer discovery, design reviews and architecture workshops, offering new ideas and not only current practice.
• Work with hardware and networking partners, such as NVIDIA, server OEMs and switch vendors, on joint designs and validations.
• Feed research results back to Product and Engineering and help shape the k0rdent AI roadmap.
• Present at industry events, webinars and partner summits, and run hands-on workshops for technical audiences.
• Write technical material, including reference architecture documents, solution briefs, blog posts and internal enablement content.

Requirements

• Bachelor's degree in computer science, electrical engineering, telecommunications or a related field, or equivalent practical experience.
• 8+ years in network engineering or architecture, including at least 3 years in data center, HPC or cloud infrastructure networking.
• Customer-facing experience as a solutions architect, pre-sales engineer, consultant or technical lead.
• Data center design: spine-leaf and Clos topologies, rail-optimised GPU fabrics, oversubscription, ECMP, and the operational problems of running thousands of nodes.
• InfiniBand: subnet management, partitioning, adaptive routing, and how NCCL and RDMA traffic behaves on the fabric.
• RoCEv2 on Ethernet: lossless, congestion-controlled designs using PFC, ECN and DCQCN, plus QoS, MTU and buffer tuning.
• Routing and overlays: BGP, EVPN-VXLAN, and VRFs for multi-tenancy.
• Hybrid and multi-cloud connectivity: interconnects, VPNs, cloud provider networking constructs, and IP address and DNS planning across sites.
• Linux networking: how the kernel and drivers handle NICs, PCI passthrough, SR-IOV, IOMMU and NUMA affinity, using tools such as iproute2, ethtool and devlink.
• Kubernetes networking: CNI plugins such as Cilium, Calico or OVN-Kubernetes, Multus and secondary networks, SR-IOV and RDMA device plugins, network policy and service exposure.
• KubeVirt, including how VM networking, passthrough and SR-IOV are set up.
• Scripting or programming in at least one language, preferably Python or Go, and working comfort with Git, CI and infrastructure as code.
• Curiosity: you investigate why a fabric behaves as it does, read specifications and source code, and keep up with changes in AI networking. You question established designs, including ones customers rely on, when a better approach exists.
• Ownership: you carry a reference architecture or proof of concept from first sketch to a published, validated result, and stand behind it.
• Initiative: you notice gaps in the product, the documentation or a customer design and act on them without waiting to be asked.
• Fast learning: you become productive quickly on unfamiliar hardware, software or customer environments.
• Business judgement: you link network design choices to cost, time to value and risk, and explain them that way to decision makers.
• Excellent written and spoken English, with documents others can build from.
• Comfort presenting to large audiences at conferences and running hands-on workshops.
• Ability to pitch the same message to network engineers and to customer executives.
• Experience in an international, distributed company across time zones and work cultures, with Solutions Architects, Partner Management and Engineering teams.
• Hands-on work with NVIDIA networking, such as Quantum InfiniBand, Spectrum-X Ethernet, ConnectX SuperNICs, BlueField DPUs, UFM and NVIDIA AI factory reference designs.
• Experience with GPU cloud, neocloud or HPC operators, or with NVIDIA Cloud Partner (NCP) style deployments.
• Bare-metal provisioning and lifecycle: Metal3, Ironic, Redfish, PXE and iPXE, and Cluster API.
• Network automation, SDN controllers, and switch operating systems such as Cumulus Linux, SONiC or vendor equivalents.
• Global load balancing and service exposure for distributed inference, including GSLB, Gateway API and ingress.
• Software-defined storage networking for AI, including NVMe-oF and GPUDirect Storage.
• Experience with the company products, MKE, k0s or k0rdent networking.
• Open-source networking or Kubernetes contributions, or a record of public talks and technical writing.
• Participation in standards or industry bodies such as the Ultra Ethernet Consortium, OCP or IETF, or published research, patents or white papers.
• Additional languages beyond English, for example a European language.

Conditions

This is a remote role open to candidates based in the United States (East Coast).

The team is spread across many time zones, so some meetings will fall outside normal local working hours.

Expect up to 25% travel for customer work, partner meetings, lab time and industry events, mainly within the EU and the US.

Pay and benefits follow the local market and the type of employment arrangement.

The company is an equal opportunity employer, and it may use automated decision-making for employment decisions, which candidates can opt out of or appeal.