Member of Technical Staff - Datacenter Networking

Prime Intellect · Remote · San Francisco · Remote ·
Fully remoteFull-timeSenior5+ years
COMPENSATION
€•• – ••k
0 applicants so far

Listed on the employer's job board (Ashby)

  • Fully remote
  • Datacenter Networking
  • GPU Cluster Focus
  • InfiniBand & RoCE
  • Senior Level

Join Prime Intellect to build the open superintelligence stack, providing infrastructure for frontier AI labs. You will own the connectivity for distributed workloads, ensuring that training and inference scale without network bottlenecks.

This position involves designing scalable topologies and automating operations for high-performance Ethernet and InfiniBand fabrics. You will collaborate with ML teams to benchmark performance and resolve complex networking issues.

What you'll do

  • Design scalable datacenter network topologies for GPU traffic
  • Operate high-performance Ethernet, RoCE, and InfiniBand fabrics
  • Automate network provisioning, validation, and upgrade procedures
  • Diagnose packet loss, congestion, and link failures across hosts
  • Benchmark end-to-end network performance with ML teams
  • Build monitoring for port health, errors, and fabric topology
  • Partner with vendors on cabling, optics, and deployment readiness

What you'll need

  • 3+ years of production datacenter networking experience
  • Strong understanding of Ethernet, TCP/IP, routing, and switching
  • Hands-on experience with InfiniBand or RoCE for GPU networking
  • Ability to troubleshoot Linux hosts, NICs, switches, and physical links
  • Experience automating network operations with Python or Ansible
  • Knowledge of leaf-spine architectures, BGP, ECMP, and VLANs
  • Familiarity with RDMA concepts, congestion control, and lossless Ethernet

Nice to have

  • Experience operating 400G/800G networks or large GPU clusters
  • Background with NVIDIA Spectrum or Quantum networking
  • Skills in NCCL performance analysis and distributed training troubleshooting
  • Knowledge of EVPN/VXLAN, SONiC, or network source-of-truth systems
  • Experience with network simulation and automated capacity planning

Stack

  • Ethernet
  • InfiniBand
  • RoCE
  • Python
  • Ansible
  • BGP
  • ECMP
  • VLANs
  • RDMA
  • Linux
  • NVIDIA Spectrum
  • SONiC

How you'll work

  • Fully remote position
  • Global availability