Domain Architect - AI Compute

World Wide Technology; Sydney; Presencial; AUD 260.000 - 380.000; Jornada completa

Descripción de la vacante

World Wide Technology, Inc. is seeking a Domain Architect, AI Compute, to own the Compute pillar across client environments. This hands-on leader will balance delivery and pre-sales SME work, comfortable in CLI and presenting to executives.

You will lead provisioning of NVIDIA NVL72, DGX SuperPOD, and related storage and networking, configure Slurm, NCCL, and telemetry, and partner with customers through workshops and discovery sessions.

Formación

• Experience with GPU compute in enterprise or HPC environments.
• Strong knowledge of NVIDIA hardware and software stack.

Responsabilidades

• Lead provisioning of NVIDIA NVL72, DGX SuperPOD, BasePOD, HGX, MGX and NVSwitch fabrics.
• Configure Slurm, Run:AI, Kueue and Volcano for multi-tenant environments.
• Implement GPU telemetry and validate performance with NCCL tests and HPC benchmarks.
• Lead Day 2 handover with identity and storage integrations.
• Support pre-sales: scope requirements, LOE estimates for SOWs, define SOE and BOM accuracy.

Conocimientos

GPU compute expertise

Pre-sales collaboration

Workshop facilitation

Client-facing

Scripting/Automation

Herramientas

NVIDIA CUDA

cuDNN

NCCL

Python

Ansible

Slurm

Base Command Manager

Mission Control

Descripción del empleo

As Domain Architect, AI Compute, you will be the primary technical authority for GPU compute fleets across client environments, owning the "Compute" pillar of the Compute-Network-Storage triad. This is a hands‑on role with a 60/40 split between delivery and pre-sales SME work, and you will be as comfortable in the CLI as presenting to C-level stakeholders.

Key responsibilities

• Lead physical provisioning of NVIDIA NVL72, DGX SuperPOD, BasePOD, HGX, MGX and Cisco AI Factory clusters
• Use NVIDIA Base Command Manager and Mission Control for provisioning, hardening and monitoring, with automated Zero Touch Provisioning
• Configure Slurm, Run:AI, Kueue and Volcano, including fair-share, MIG quotas and preemption for multi-tenant environments
• Implement GPU telemetry (DCGM) and validate performance with NCCL-tests, HPL and HPCG
• Lead Day 2 handover, integrating identity providers and storage backends
• Support pre-sales: scope requirements, estimate LOE for SOWs, define the SOE, own Compute BoM accuracy and run client discovery workshops

About you

Current Australian work rights; based in Sydney or Melbourne

Deep knowledge of NVIDIA Hopper, Grace-Hopper, Blackwell and Grace-Blackwell

Proven mastery of NVL72 rack‑scale integration and NVSwitch fabrics

Expert knowledge of CUDA, cuDNN and NCCL

Python and Ansible for automation

Proven hands‑on experience building and handing over GPU clusters

Strong client-facing and workshop facilitation skills