Applied Researcher – Network Expert

Overview

APPLIED RESEARCHER – NETWORK EXPERT

Location: Hybrid | Bellevue, WA (downtown)

ABOUT THE OPPORTUNITY

Our client is seeking a Network Expert to join an Applied Research organization focused on the future of large-scale AI infrastructure.

This role is designed for a networking expert with deep experience in GPU-based computing environments and high-performance inter-GPU networking. As AI workloads increasingly rely on thousands of GPUs operating as a single logical system, the network connecting those GPUs becomes a critical component of overall infrastructure performance and scalability.

You will serve as a horizontal subject-matter expert, partnering with engineering and infrastructure leaders to evaluate technologies, architectures, and industry developments. You will help the organization understand where GPU networking is headed and what those developments mean for infrastructure, architecture, and investment decisions.

WHAT YOU'LL DO

• Hold the company's view of where AI networking is going.
• Track vendor and hyperscaler roadmaps, research, standards work, and the startup and venture landscape across scale-up, scale-out and scale-across fabrics, topology and optics, collectives and the software above the fabric, operations and telemetry, and tenant-facing capability. Right now that means questions like how fast Ultra Ethernet displaces the RoCEv2 fabrics most clusters actually run, where the scale-up domain should end now that NVLink has open challengers, what co-packaged optics does to power per port, and how to network a cluster that no longer fits in one building. Those specific questions will have changed within a year — holding the current version of them is the job.
• Formulate and validate the product and engineering thesis. Turn that view into a defensible position on what we build, buy, or partner for in the fabric, pressure-tested against measured cluster performance, isolation requirements, and cost per port — and say so plainly when the evidence does not hold up.
• Own the company's fabric position: where the scale-up and scale-out boundary sits for our workloads, which transport we bet on and when, and what good telemetry and observability look like so fabric problems are diagnosable rather than inferred. A fault that restarts a long training job is a direct cost, not an availability statistic.
• Help finance to formulate the numbers: cost per port, optics and cabling economics across pluggables and co-packaged options — reach, power draw, failure rates at scale — and the performance we can actually substantiate against what vendor benchmarks claim.
• Own tenant-facing network capability for GPU-as-a-service: multi-tenant isolation and its performance cost, storage traffic alongside GPU traffic, and what we can commit to contractually — including an honest read of where we are undifferentiated against peers.
• Make the work land commercially. Support sales and delivery in demanding customer conversations about cluster performance, feed product and go-to-market with what we can offer at what performance and price, and provide technical diligence on network vendors, partners, and prospective tuck-in targets. Work closely with the Data Center Expert where the fabric meets the physical plant.
• Deep professional experience in networking at scale,
• Hands-on experience with high-performance GPU or HPC fabrics at current generations,
• Demonstrated experience debugging real collective communication performance problems in production
• Fluency with the landscape you would be scanning: the switch, NIC, and optics vendors, the standards bodies and consortia, and the startups attacking the fabric layer — and a view on which of them matter. Expect to be asked what you think is currently overhyped, and why.
• Demonstrated ability to do research in the applied sense: taking an open question, investigating it

WHAT WE'RE LOOKING FOR

REQUIRED QUALIFICATIONS

Preferred Qualifications