Job Summary
Strong hands-on experience with BGP (eBGP underlay, BGP unnumbered, ECMP) in leaf-spine/Clos data center fabrics.Design and deployment of VXLAN-EVPN overlays, IRB, and multi-tenancy with VRFs.Basic conceptual knowledge of InfiniBand, RDMA/RoCEv2, and lossless Ethernet tuning (PFC, ECN, DCQCN).Familiarity with the NVIDIA Spectrum-X Ethernet platform, ConnectX SuperNICs, and BlueField DPUs.Good knowledge of Linux and the ability to quickly pick up Cumulus Linux.Understanding of GPU cluster network design based on DGX reference architectures.Fabric monitoring and troubleshooting using streaming telemetry and Linux networking tools, down to 400G/800G optics and cabling.Good understanding of rail-optimized architecture and the separation of frontend and backend fabrics in GPU clusters for east-west (E-W) and north-south (N-S) traffic flows.Good hands-on experience with Linux and Kubernetes.Good working understanding of HPC environments.
Key Responsibilities
Hands-on experience with NVIDIA DGX and GPU management/administration.Good understanding of AI training and inference.Deep hands-on experience with multiple network operating systems, such as Cumulus Linux, SONiC, IOS, Junos OS, and EOS.Proven ability to architect, deploy, troubleshoot, and optimize large-scale production clusters.Advanced understanding of congestion control, traffic engineering, topology design, telemetry, and network performance optimization.Extensive experience with automated provisioning, shell scripting, Python, Ansible, Salt, APIs, and CI/CD pipelines.Strong understanding of the interactions among networking, GPU computing, distributed storage, and HPC workloads.Ability to create HLD/LLD documents and validate designs before production rollout
Skill Requirements
Kubernetes, networking, ansible, HPC