Senior Enterprise Network Engineer

Nebius is building a full-stack AI cloud that combines hyperscaler scale with supercomputer performance. GPU clusters only deliver if the fabric never becomes the bottleneck. You will own enterprise-grade network design and operations so training and inference stay lossless, predictable, and production-ready for customers worldwide.

Performance Objectives

• Design, qualify, and bring up multi-tier CLOS fabrics (EVPN/VXLAN, 400G/800G-ready) so new GPU clusters pass acceptance and hit target RoCE/InfiniBand latency and loss within the first production window.
• Drive DCI, backbone, and enterprise interconnect (BGP, MPLS/SRv6, PNI/transit) so regional sites stay dual-homed, policy-clean, and ready for follow-the-sun operations.
• Diagnose and close Sev-1 fabric issues (PFC/ECN, ECMP imbalance, NIC/switch errors) under real training and inference load, restoring SLAs without repeating the same class of incident.
• Automate Day-0 through Day-2 (OOB, config, telemetry, NetBox) so greenfield colo and campus builds are repeatable and handoff-ready for global peers.
• Partner with HPC, hardware vendors, and enterprise customers to qualify NICs, optics, and tunings that measurably improve NCCL/MPI job efficiency.
• Write runbooks and onboard the next engineers so regional ownership scales without heroics.

Environment & Resources

You work with core architecture, backbone, and platform teams in Amsterdam and the US, plus hardware partners (NVIDIA, Mellanox, and others). Stack spans Juniper, Arista, Cisco, NVIDIA/Cumulus, Linux, and programmable pipelines. High autonomy, global on-call, site visits as needed. Report into network engineering leadership with real ownership of fabric outcomes.

Essential Qualifications

• Hands-on production ownership of scale DC or SP networks: BGP, IS-IS/OSPF, EVPN/VXLAN, ECMP, dual-stack IPv4/IPv6.
• Proven CLOS diagnostics: packet analysis, interface errors, QoS, RoCEv2 (PFC, ECN/DCQCN) or InfiniBand behavior under load.
• Comfort from console on unconfigured OOB gear through multi-terabit edge policy (Junos, EOS, IOS-XR, or equivalent).
• Ability to self-direct in greenfield, high-velocity environments and document clearly under pressure.
• Professional English; overlap with CEST and US time zones for critical changes.

Role Selling Points

• Built by engineers, for engineers—own the hard problems across compute, storage, and networking for the AI economy.
• Nasdaq-listed (NBIS), Amsterdam HQ, global GPU footprint and real customer impact at scale.
• Trust, ownership, and a fast-moving culture that ships clusters, not slideware.

Apply if you want the fabric that actually runs the next generation of AI. Authorized to work in the country of application. Nebius is an equal opportunity employer.