Principal/Staff HPC Network Engineer

About the role

GPU clusters are some of the most performant computers on the planet. Even smaller clusters by today's standards would have ranked in the TOP500 five years ago. Our infrastructure team is responsible for architecting and deploying new clusters around the world and keeping them running smoothly. You'll participate in on-call rotation, fix issues when they arise, and lean into automation to enable deployments at scale. We're a small but ambitious team so you'll be an early contributor helping to shape culture, mentor junior engineers, and learn from our customers. About You You will have 10+ years of experience with hands-on management or architecture with network for at least one GPU cluster in the past (ideally a cluster with >1k GPU's but not required) You deeply understand the fundamentals of Ethernet (RoCEv2) and/or InfiniBand networks in CLOS/fat-tree topologies You have built HPC network architectures (eBGP, fat-tree, VXLAN, MCLAG, etc.) The idea of implementing zero-touch provisioning for a large multi-layer network excites you (you embrace automation) You appreciate, value and generate good documentation You have the ability and willingness to mentor junior engineers You're open to coming in to our San Francisco office 3-4 days per week Some Nice to Haves You understand data center concepts including power, cooling and how to engage with colo providers You have experience with Linux systems administration including managing kernel drivers and tuning the network stack You have experience with Linux virtualization (KVM, QEMU, libvirt, etc.) You've had exposure to containers and Kubernetes operators

Benefits

Additional Information

We're building the company which will de-risk the largest infrastructure build-out in history. When people finance GPU clusters, the datacenters housing them, and the infrastructure powering them, they need "offtake" - meaning someone has signed a contract to lease the cluster for a period of time before its even built. Financing a GPU cluster is inherently risky, since margins are thin and volumes are huge. Lenders don't want to take on the risk that cluster developers can't repay their loan, and cluster developers really don't want to risk not selling their cluster. As a result, risk is offloaded to the customer using fixed-price long-term contracts. If you don't mitigate this customer risk, there's a bubble. This isn't SaaS anymore - application layer companies sign multi-year contracts for computer and inference, but sell to customers on monthly subscriptions. If you mess up a purchase, it's game over: a minor shift in your revenue growth rate might mean the difference between profit or bankruptcy. But what if companies could exit their contract by selling it back to the market? Otherwise, as AI scales, compute only becomes available to folks who can effectively take on that risk. A 2-person startup in a San Francisco Victorian can't realistically sign a 5-year take or pay contract on $100m supercomputers. But they may be able to buy the month of liquidity that someone else sold back. So that's what we make: a liquid market for GPU offtake.

Your Match

How well this role fits your profile.

Company Intel

What employees say

Worked at sfcompute? Share your experience