GPU Infrastructure Engineer

NODE ONE SDN. BHD.; Kuala Lumpur; On-site; MYR 134,000 - 223,000; Full time

Benefits offered by this job

Travel allowance

Global-project exposure

Career growth

Job summary

NODE ONE SDN. BHD. is seeking a hands-on GPU Infrastructure Engineer to design, deploy, optimise and operate GPU networks, storage, and performance layers for large-scale AI compute clusters.

You will own the technical backbone enabling fast, stable, and cost-efficient training and inference, with regular international travel to deploy and support on-site.

Qualifications

• 3+ years hands-on experience with GPU/HPC clusters and multi-node delivery/operations.
• Solid GPU networking: InfiniBand/RoCE, NVLink/NVSwitch, Ethernet fabrics, NCCL.
• Experience profiling/optimising CUDA workloads and large-model training.
• Familiar with parallel storage and high-throughput data pipelines (Lustre, Ceph, NVMe-oF).
• Strong Linux administration and scripting (Python/Bash/Go); containerisation with Docker/Singularity; Slurm and/or Kubernetes.

Responsibilities

• Design, deploy, optimise and operate GPU cluster fabrics and storage for AI compute.
• Plan cluster topologies, cabling, and capacity planning with site teams.
• Troubleshoot RDMA, NCCL, and performance bottlenecks across distributed jobs.
• Profile and tune workloads (CUDA, Nsight) for training and inference; maximise utilization.
• Drive cost-performance and continuous optimisation for customer workloads.
• Travel internationally 30–50% for deployment and support.
• Build docs and runbooks; transfer knowledge to regional teams.

Skills

GPU HPC clusters

InfiniBand networking

NVIDIA NVLink/NVSwitch

NCCL tuning

Linux administration

Shell scripting

Python

Go

Docker

Kubernetes

Slurm

NVIDIA tools

Education

Bachelor's degree in CS/EE

Tools

Docker

Singularity

Kubernetes

DCGM

Prometheus

Grafana

fio

Job description

The Role

We are looking for a hands-on GPU Infrastructure Engineer to design, deploy, optimise and operate the GPU networking, performance, and storage layers of our AI compute clusters. You will own the technical backbone that makes large-scale model training and inference fast, stable and cost-efficient. Because our projects and customers are global, this role requires regular international travel to deploy, troubleshoot and support on-site.

Key Responsibilities

Design, deploy and maintain high-performance GPU cluster fabrics — InfiniBand (NDR/XDR), RoCE v2, NVLink / NVSwitch, and 100/200/400 GbE Ethernet.

Plan cluster topologies for multi-node training, and commission rack-level and cluster-level cabling with the DC / site team.

Diagnose and resolve network bottlenecks, RDMA issues, and NCCL collective-communication stalls across distributed jobs.

Our newest clusters are built on NVIDIA B300 (Blackwell) GPUs — experience with B300 scale-up (NVLink/NVSwitch) and scale-out (InfiniBand NDR/XDR, Quantum-X800) networking is strongly preferred.

Profile and tune GPU workloads (CUDA, Nsight) for training and inference; maximise GPU utilisation and cost-performance.

Optimise distributed training — PyTorch DDP/FSDP, DeepSpeed, Megatron — covering data / tensor / pipeline parallelism and NCCL tuning.

Benchmark clusters, build performance baselines, and drive continuous optimisation for our customers' model workloads.

Design and operate high-throughput parallel storage for AI — Lustre, BeeGFS, WekaIO, GPFS, Ceph, NVMe-oF.

Optimise data pipelines and checkpointing to eliminate I/O stalls during large-model training.

Own capacity planning, tiering (NVMe/SSD/object), and storage performance tuning.

D. Global Deployment & Customer / Project Support

Travel internationally to customer sites and project locations for cluster deployment, PoC, acceptance testing and on-site troubleshooting.

Serve as the technical point of contact for GPU/network/storage during project delivery and post-deployment support.

E. Documentation & Knowledge

Build runbooks, deployment standards, and best-practice playbooks; transfer knowledge to local and regional teams.

Requirements

Bachelor's degree or above in Computer Science, Electrical / Electronic Engineering, or a related field.

3+ years hands-on experience with GPU / HPC clusters (NVIDIA A100 / H100 / B300, DGX or equivalent, multi-node) and cluster delivery / operations.

Solid GPU networking: InfiniBand / RoCE RDMA, Mellanox switches & NIC deployment and tuning, NVLink / NVSwitch, Ethernet fabrics, and NCCL communication principles.

Practical GPU tuning: CUDA profiling, NCCL (env-var tuning and communication-bottleneck diagnosis), and real multi-node / multi-GPU large-model training optimisation.

AI storage: parallel file systems (Lustre preferred, then Ceph / JuiceFS), NVMe-oF, high-throughput data pipelines, checkpoint optimisation, and storage IO benchmarking (fio / vdbench / ior).

Strong Linux administration (CentOS / RHEL / Ubuntu) and shell scripting (Python / Bash / Go); containerisation with Docker and Singularity; experience with Slurm and/or Kubernetes (GPU Operator, Device Plugin, Volcano).

Familiar with the NVIDIA stack: DCGM, MIG, NVLink / NVSwitch; able to read nvidia-smi and dcgmi logs for diagnostics.

Fluent working English (written and spoken).

Chinese language ability (Mandarin) — required for coordination with group HQ and cross-border teams.

Willing and able to travel internationally (est. 30–50% of time) to serve global customers and projects.

Preferred / Nice to Have

Hands-on experience with NVIDIA B300 / Blackwell GPU cluster networking — node-scale NVLink/NVSwitch and cluster-scale InfiniBand (NDR/XDR, Quantum-X800) or Spectrum-X Ethernet for scale-up / scale-out.

NVIDIA certifications (e.g. NCP / NCU) or Mellanox / NDR networking certification.

Experience with liquid-cooled, high-density (50kW+) data-centre deployments.

Familiarity with AI PaaS / inference serving stacks (vLLM, TensorRT-LLM, TRT-LLM).

Monitoring & observability: DCGM, Prometheus, Grafana, training-job observability.

Prior experience in a hyperscale / AIDC or cloud-GPU environment.

What We Offer

Competitive salary (indicative range RM 12,000 – RM 20,000/month, commensurate with experience — to be confirmed by HR).

Travel allowance and global-project exposure across Southeast Asia and beyond.

The chance to build cutting-edge GPU infrastructure from the ground up at a fast-scaling AI group.

Clear career growth path within the regional engineering organisation.