GPU Networking Engineer - RoCE & InfiniBand

kyndryl; Setiabudi; Presencial; IDR 350.000.000 - 550.000.000; Jornada completa

Descripción de la vacante

Kyndryl is seeking an RoCE/InfiniBand Network Engineer to troubleshoot and implement changes for high-speed GPU fabrics. You will work with RoCE and InfiniBand environments, owning service restoration and escalating complex fabric conditions to SMEs.

You will analyze errors, manage L1- to L2-grade incidents, and collaborate with compute teams for optimal NCCL workloads. A strong focus on VLAN-less, lossless networks in data centers.

Formación

• 3–5+ years of data-center network engineering experience, with hands-on high-speed fabric work.
• Production RoCEv2/RDMA or InfiniBand experience; GPU/HPC environments preferred.
• Hands-on switch CLI, optics/cabling, NIC/HCA diagnostics and performance counters.
• Knowledge of DCB, PFC, ECN, QoS and lossless Ethernet for RoCE.
• Familiarity with UFM and subnet-management for InfiniBand.
• Experience with leaf-spine, BGP/EVPN and high-speed fabrics (100/200/400 Gb+).
• Formal incident/change/RCA experience.

Responsabilidades

• Troubleshoot switch, HCA/NIC, port and optic issues using CLI and fabric tools.
• Analyze link errors, drops, flaps, speed, FEC and RDMA reachability.
• Operate InfiniBand/UFM or Spectrum-X/RoCE management tooling within policy.
• Troubleshoot RoCE: DCB, PFC, ECN, QoS and common lossless-network issues.
• Troubleshoot InfiniBand: link/subnet/fabric conditions and HCA state.
• Participate in BGP/EVPN and fabric change implementations per site architecture.
• Correlate network symptoms with multi-node GPU/NCCL workloads and engage compute teams.
• Perform firmware/OS changes, cable replacements and post-change validation.
• Own SLA/ticket progression; contribute to RCA; mentor L1 engineers.

Conocimientos

Data-center networking

RoCEv2/RDMA

InfiniBand

switch CLI

DCB/PFC/ECN

QoS

UFM

subnet-management

leaf-spine

BGP/EVPN

NVIDIA Spectrum-X/ConnectX

NCCL workload correlation

Herramientas

NVIDIA Spectrum-X

ConnectX

Fabric-management tooling

Descripción del empleo

Kyndryl is seeking an RoCE/InfiniBand Network Engineer to troubleshoot and implement changes for high-speed GPU fabrics. You will work with RoCE and InfiniBand environments, owning service restoration and escalating complex fabric conditions to SMEs.

You will analyze errors, manage L1- to L2-grade incidents, and collaborate with compute teams for optimal NCCL workloads. A strong focus on VLAN-less, lossless networks in data centers.