Network Engineer/AI Fabrics Architect, AI Solutions Lab - Santa Clara, CA, United States

Originally posted 16 October 2026 by the employer.

Design, tune, and validate RoCE-based AI fabrics and Marvell's Teralynx switching under production-like conditions.

About the role

This role involves validating Marvell's Teralynx switching and AI fabrics under production-like conditions across compute, networking, and memory fabrics. The team runs AI benchmarking experiments mirroring real customer workloads and operates GPU/HPC lab clusters.

What you'll do

• Own Teralynx-based switch configuration, SONiC deployment, and lab fabric design.
• Design, tune, and validate RoCE-based AI fabrics for scale-up and scale-out workloads.
• Architect rail-optimized, leaf-spine, fat-tree, and multi-plane fabric topologies for distributed training and inference clusters.
• Configure and optimize ECN, PFC, UEC, congestion-control behavior, and telemetry pipelines.
• Analyze GPU collective communication patterns (all-reduce, all-gather, broadcast, reduce-scatter) and their impact on east-west traffic and fabric topology design.
• Debug packet loss, traffic hotspots, head-of-line blocking, and queue behavior using counters and fabric telemetry.

What you'll need

• Expertise in data center, backbone and internet protocols and technology.
• Experience with RoCE and AI-fabric tuning in high-performance computing or data center environments.
• Working understanding of GPU collective communication patterns (all-reduce, all-gather, reduce-scatter) and how they shape network topology decisions.
• Strong hands-on knowledge of switch automation, telemetry collection, and counter analysis.
• Experience with SONiC, SAI, network SDKs, and programmable switching platforms.
• Familiarity with Ethernet-based AI fabrics and emerging AI interconnect architectures, including high-speed Ethernet (200G/400G/800G/1.6T).
• Experience deploying telemetry for network fabrics using SNMP, gNMI, or gNXI.

Nice to have

• Experience with containers and orchestration frameworks such as Kubernetes.
• Experience working with CI/CD pipelines to manage lab infrastructure.

Skills: Teralynx-based switch configuration, RoCE-based AI fabrics, GPU collective communication patterns, SONiC, Ethernet-based AI fabrics