Senior Network Engineer

Senior Network Engineer Infrastructure operations · shared across customers Reports to: Director, Network Engineering Location: Remote (US) or Pleasanton, CA (hybrid) Department: Infrastructure & DC Operations / Network Engineering Position summary The Senior Network Engineer designs, deploys, and operates the high-performance networking fabric supporting GPU clusters. This includes InfiniBand and RoCE fabrics for training workloads, customer-facing connectivity, and the wide-area network that connects STN sites and customer environments. Key responsibilities • Design and configure InfiniBand or RoCE fabrics optimized for GPU training and distributed inference • Configure and operate switching, routing, and customer VLAN/VRF/VPC architectures • Manage BGP peering, public IP space, anycast, and DDoS protection • Design customer connectivity including cross-connects, dedicated links, VPN, and SD-WAN • Maintain network automation, configuration management, and source-of-truth tooling • Coordinate with the NOC on network monitoring, alerting, and runbook authoring • Troubleshoot complex network issues across layers 1 through 7 • Maintain network documentation, diagrams, and operational runbooks • Drive network capacity planning aligned to fleet growth and customer commitments • Support security and compliance audits including SOC 2 and customer security reviews Required qualifications • 7+ years in network engineering with data center or service provider experience • Deep expertise in InfiniBand or RoCE (RoCEv2), including congestion control and NCCL tuning • Strong knowledge of BGP, OSPF, MPLS, VXLAN, and EVPN • Hands-on experience with Arista, NVIDIA Mellanox/Spectrum, or Cisco platforms • CCIE, JNCIE, NCIE, or equivalent advanced certification strongly preferred Preferred qualifications • GPU cluster networking experience at multi-thousand-GPU scale • SDN and automation skills (Ansible, Python, Nautobot, or Netbox) • Multi-site WAN and peering experience including IX participation Familiarity with NVIDIA Cumulus, SONiC, or open networking stacks