AI Factory Network Architect

Runware is the AI inference platform that builds and runs its own infrastructure, end to end. Our Sonic Inference Engine has served more than 20 billion requests for over 1 million developers and 500 million end users, with 400,000+ models behind a single integration, at up to 10x lower cost per token than traditional providers.

We are now building the AI Factory that runs it. The Sonic Inference Pod packs 1 MW of custom-designed compute into a 20-foot container with closed-loop, water-free cooling, sited wherever power exists. Pods are live in the US and Europe, and we are scaling to dozens of pods and tens of thousands of GPUs, all operated as one distributed fabric that reroutes traffic the moment a pod goes offline (announcement).

The network is what turns those pods into a supercomputer. That is the job.

The role

As AI Factory Network Architect, you will own the network architecture of Runware's AI supercomputer, from the NVLink domain inside a server to the optical links between pods on two continents. You will set the design, build the first systems with your own hands, and grow the team that runs them.

This is a hands-on leadership role, not a slide-deck role. You will specify the fabrics, validate them on real hardware, debug the first congestion collapse at 3 a.m., and then write the standard so it never happens again. You will work directly with our hardware, platform, and inference engineering teams, with NVIDIA and our optical and switching vendors, and with the site teams deploying pods.

Success means inference traffic flows across tens of thousands of GPUs at line rate, pods join the fabric in days rather than weeks, and a failure anywhere in the network is invisible to the developer calling our API.

What you'll do

Own the end-to-end network architecture for the AI Factory: the scale-up fabric (NVLink and NVSwitch domains), the scale-out fabric (InfiniBand and RoCE), the storage and management networks, and the pod-to-pod and inter-site backbone.

Design the optical layer: transceiver and cable selection, co-packaged and linear-drive optics evaluation, DWDM for pod-to-pod and metro links, and the link budgets and reliability targets behind them.

Define the reference network design for the Sonic Inference Pod so every new pod ships with a validated, repeatable fabric and joins the global network in days.

Build and bring up the first fabrics yourself: rail-optimized topologies, subnet manager and routing configuration, adaptive routing, congestion control, QoS, and GPUDirect RDMA tuning, measured against inference throughput and tail latency rather than synthetic benchmarks.

Work with the inference engine team so the Sonic Inference Engine's traffic patterns, from tensor- and pipeline-parallel collectives to KV-cache transfer and multi-node serving, drive topology and buffer decisions, not the other way round.

Run capacity planning and the vendor roadmap with NVIDIA, switch and optics suppliers, and carriers; negotiate technical specifications and lead evaluations of new fabric generations.

Set the operating standard: telemetry, link-health monitoring, failure-domain design, automated fault isolation, and runbooks that let a small team operate tens of thousands of ports.

Hire and lead a team of network engineers as the footprint grows, and act as the final technical escalation for fabric incidents across all sites.

What we're looking for

15+ years designing, building, and operating high-performance networks, with a significant part of that in HPC or AI supercomputers at the scale of thousands of GPUs or more.