Network Architect

Network Architect

Runware · Palo Alto, United States (Hybrid) · Full-time · Posted 2026-11-04

Workplace: hybrid

Department: Infrastructure

Description

About Runware

Runware is the AI inference platform that builds and operates its own infrastructure, end to end. One API gives developers access to image, video, audio, 3D and language models, and our Sonic Inference Engine runs them on hardware we designed ourselves: custom servers, custom boards and a closed-loop, water-free cooling system, packed into Sonic Inference Pods — roughly 1 MW of liquid-cooled GPU compute in a 20-foot container that can be sited wherever power is cheap and available.

This is why infrastructure at Runware is the product, not a cost center. Our edge is that we run inference 30–90% cheaper per GPU-hour than our peers, and that edge is earned in the data center: in how fast we can stand up a new site, how tightly we run our fleet, and how little downtime our customers ever see. More than 20 billion requests have run through the platform for over a million developers, with customers such as Wix, Quora, Freepik and Higgsfield building production applications on top of it.

We raised a $50M Series A led by Dawn Capital, with Comcast Ventures, Speedinvest, Insight Partners and a16z speedrun participating, and we are now building out a global network of pods across the US and Europe. We need someone to own how that network is operated.

The Role

As Network Architect, you'll own the design, build, and operation of the network that moves every inference request through Runware: the GPU cluster fabric inside our data centers, the backbone and edge that deliver results to customers, and the cloud and hybrid connectivity that ties partner capacity into one platform. Runware's product is latency. The network is where a large share of that latency, and of our cost per request, is decided, and today there is no single owner for it. You'll be that owner.

This is a senior, hands-on architecture role with significant authority and no direct reports to start. You'll report to Risto Haukioja, Head of AI Infrastructure. You'll set the network strategy, design and build the systems yourself, define how they are operated and capacity-planned, and lead through design authority across Infrastructure, Platform, and ML Engineering. You've run high-performance teams before, and you'll shape the standards and hiring plan for network engineering as the footprint grows. We expect you to use AI tooling as a core part of how you work, so that one architect can produce what a small team used to.

What you'll do

• Own the network architecture end to end. Define the target architecture and multi-year roadmap for Runware's data center fabric, backbone and edge, and cloud and hybrid connectivity, and make the design decisions that trade off latency, throughput, cost, and resilience.
• Design and build the GPU cluster fabric. Architect spine-leaf fabrics for multi-node inference and training: 400G/800G Ethernet, RoCEv2 and InfiniBand, lossless transport, congestion control, and topology choices that keep east-west traffic from becoming the bottleneck.
• Deliver inference to customers fast and reliably. Own the backbone and edge: multi-region interconnect, transit and peering strategy, anycast and traffic steering, DDoS and edge security posture, and the path that gets every response to the customer with as few milliseconds as possible.
• Connect cloud, bare metal, and partner capacity into one platform. Design VPC, VPN, and private interconnect across hyperscalers, bare-metal providers, and colocation sites so new capacity can be brought online quickly and operated consistently.
• Build network operations, not just networks. Define how the network is monitored, changed, and repaired: telemetry and observability, change management, incident runbooks, and on-call practices that hold up at scale. Stand up what doesn't exist yet.
• Run capacity planning. Build the models that forecast bandwidth, port, and interconnect needs from product growth, and turn them into procurement and build plans with lead times that match the business, not the other way round.
• Automate the network. Treat network configuration as code: IaC, intent-based provisioning, automated validation and rollback, and telemetry-driven remediation, so the network can grow without a proportional growth in people.
• Accelerate your own output with AI. Use AI tools throughout your work (design exploration, configuration generation, log and telemetry analysis, documentation, automation code) and set the standard for how the wider infrastructure team does the same.
• Lead through design authority. Set network standards, review designs across Infrastructure, Platform, and ML Engineering, mentor engineers, and define the hiring plan and operating model for network engineering as the footprint grows.
• Own vendor and carrier relationships. Evaluate and select hardware, optics, carriers, and interconnect partners, and negotiate for performance and cost.

Requirements

• 15+ years in network engineering and architecture, including at least 5 years as the design authority for large-scale production networks at a cloud provider, hyperscaler, CDN, HPC/AI infrastructure company, or carrier.
• Hands-on. You still design, configure, and troubleshoot networks yourself, and you've built data center, backbone, and cloud networks from the ground up, not just inherited them.
• Deep, current expertise in modern data center networking: spine-leaf Clos fabrics, EVPN/VXLAN, BGP at scale, 400G/800G Ethernet, RoCEv2 and lossless Ethernet (PFC, ECN, DCQCN), and InfiniBand for GPU clusters.
• Strong backbone and edge experience: multi-region interconnect, transit and peering, anycast, traffic engineering, and DDoS mitigation, with a track record of measurably reducing customer-facing latency.
• Cloud and hybrid networking across at least two hyperscalers plus bare-metal or colocation environments: VPC design, private interconnect (Direct Connect, ExpressRoute, Cloud Interconnect or equivalent), and VPN at scale.
• Experience designing and running network operations: observability, change management, incident response, and on-call, with the judgement to know what to build first when starting from little.
• Proven capacity planning: you've built forecasting models that turned product growth into bandwidth, port, and interconnect plans, and you've run the procurement cycles that followed.
• Network automation as a working practice, not an aspiration: Python or Go, IaC for network config, CI-style validation, and streaming telemetry.
• You use AI to accelerate your own output. You can show how AI tooling is part of how you design, write configuration and automation code, analyse telemetry, and document, and you expect it to multiply what you ship.
• Experience leading high-performance engineering teams: you've hired, set standards, and run teams that delivered under pressure, and you lead well through influence when you have no direct reports.
• Strong ownership and comfort in a fast-moving, ambiguous environment where you identify the problems, create clarity, and drive the work to done.
• Based in, or with full working overlap with, US Pacific time.

Strong Differentiators

• Experience networking large GPU clusters for AI inference or training (NVIDIA Spectrum-X, Quantum InfiniBand, NCCL tuning, GPUDirect RDMA), and an opinion on Ethernet versus InfiniBand backed by production numbers.
• Hands-on with modern network operating systems and disaggregated hardware: SONiC, Cumulus, Arista EOS, Junos, or NVIDIA Cumulus/Spectrum platforms.
• Experience with Ultra Ethernet, 800G and 1.6T optics, co-packaged optics, or other emerging fabric technologies.
• Experience running peering and transit for a latency-sensitive product, including ASN management, route optimisation, and IXP presence.
• Experience with eBPF, DPDK, SmartNICs/DPUs, or kernel-bypass networking on the host side.
• Prior work at a startup or scale-up where you built the network function from zero.
• Public contributions: RFCs, open-source network tooling, conference talks, or published post-mortems.

Benefits

We're a remote-first collective, meeting in person twice a year to plan, brainstorm, celebrate wins, and enjoy some face-to-face time. We have core hours for cooperative working and calls, but outside of that your calendar is yours. Work the hours that let you perform at your peak while also building a healthy life.

Our release cycles are fast and intense, but they're followed by real downtime. After big pushes we expect the team to unplug, recharge, and come back ready and stronger than ever for the next leap.

• Generous paid time off: vacation, sick days, public holidays
• Meaningful stock options: share in the upside you create
• Flexible hours: own your schedule outside core collaboration blocks
• Family leave: paid maternity, paternity, and caregiver time
• Company retreats: twice-yearly gatherings in inspiring locations

Apply

Apply at Runware

---

Powered by Workable