Senior Cloud Infrastructure and Network Operations Solutions Architect
Pubblicato il 04-10-2026 - NVIDIA in Montà
NVIDIA is looking for a Senior Network and Fabric Operations Solutions Architect to join its NVIDIA Infrastructure Specialist Team.
È pronto/a a candidarsi? Si assicuri di aver compreso tutte le responsabilità e i compiti associati a questo ruolo prima di procedere.
Academic and commercial organizations around the world are using NVIDIA products to redefine deep learning and data analytics, and to power next-generation data centres.
Join the team building and advising on the networks behind many of the largest and fastest AI/HPC systems in the worldWe are looking for someone who combines deep networking expertise with strong consulting and communication skills.
This role will engage directly with customers, partners and cross-functional teams to assess, architect and operate the compute, storage and scale-up fabrics that underpin large GPU estates—serving as both a trusted advisor and a hands-on technical leader.
You will own the fabric side of NVIDIA’s Cloud Partner (NCP) operating model across the full Day 1 to Day 2 lifecycle:
from InfiniBand/UFM and Ethernet validation at handover, through NVLink partition operations, to a production-stable fabric running at maximum goodput.
NCP estates are open-source-first and heterogeneous—Cumulus Linux, SONiC, Spectrum-X, DOCA/DPU services, upstream Kubernetes and Prometheus/Grafana—so this role is deliberately tool-agnostic: you will meet each partner on the fabric they actually run rather than on a single proprietary product.What You’ll Be Doing:Own Day 2 fabric operations across NCP fleets: NVLink/NVSwitch partition management (NMX-C / NMX-M) on GB200/GB300 NVL72-class systems, maintenance-partition isolation, and safe partition-change workflows for multi-tenant estates.Own the switch software and firmware lifecycle—Cumulus Linux, SONiC and switch-OS upgrades, CPLD and field-notice rollout campaigns—with staged, tested upgrade plans that keep partner fabrics off stale releases without disrupting production workloads