Note: The job is a remote job and is open to candidates in USA. Realm is partnering with a profitable, self-funded cloud infrastructure company that operates a global data centre footprint and is expanding its GPU and AI infrastructure. The Senior Network Engineer will own GPU cluster networking, lead fabric buildouts and migrations, resolve critical performance issues, automate network operations, and mentor engineers in a 24x7x365 environment.
Responsibilities
- Own the fabric behind the GPU clusters. When a customer's training run stalls on latency or a fabric won't hit line rate, you're the person who finds out why and fixes it
- Lead the design and deployment of GPU cluster buildouts and migrations across the global network
- Serve as senior escalation for critical network incidents: root cause analysis, cross-functional coordination, permanent remediation
- Work directly with strategic customers to onboard GPU deployments and solve real performance problems around latency, jitter and reachability
- Champion automation. Build the tooling that kills manual toil rather than living in it
- Mentor the engineers around you and raise the team's baseline through documentation, pairing and knowledge sharing
- On-call rotation and incident lead during critical events, in a genuine 24x7x365 operation
Skills
- - 7+ years in network operations or network engineering, with clear progression of responsibility
- - Hands-on experience with GPU cluster networking: RDMA/RoCEv2 fabric config, lossless Ethernet design, and congestion control (ECN, PFC, DCQCN) in multi-tenant environments
- - Solid grasp of high-performance topologies for AI/ML: leaf-spine / fat-tree fabrics built for collective communication patterns (all-reduce, all-gather), and the interconnects behind them (NVLink, NVSwitch, InfiniBand)
- - Expert-level BGP on internet-scale networks: routing policy, peering, RPKI, ECMP
- - Strong with VxLAN/EVPN in large-scale data centre fabrics
- - Strong Linux (Debian a plus): kernel routing, nftables/iptables, and routing suites like FRR or Bird
- - Comfortable across more than one vendor and NOS, e.g. Juniper (Junos), Cisco (IOS-XR/NX-OS), Nokia (SR OS), Cumulus, SONiC
- - Real Python and Ansible ability for repeatable change and health data collection
- Fully remote within the US
- Flexible schedule to cover scheduled maintenance, which can include some weekends and holidays
- - Exposure to GPU/AI platforms and vendors: NVIDIA (Spectrum-X, Mellanox), Arista (7800R3), Nokia (7750 SR)
- - Familiarity with cluster management tooling (Slurm, Kubernetes with GPU device plugins) and GPU health monitoring (nvidia-smi, DCGM)
- - CCNP, JNCIS or equivalent demonstrable depth
Benefits
- Fully remote within the US.
- Flexible schedule to cover scheduled maintenance, which can include some weekends and holidays.
Company Overview
Company H1B Sponsorship