Senior Product Architect, Storage

About the role

Reference architecture work for disaggregated inference, foundation model training and agentic pipelines, done jointly with storage partners. Success means validated, production-ready designs that hold up at datacenter scale.

At a glance

What you will do

• Design full-stack reference architectures for disaggregated inference, large model training and agentic AI, built with storage partners
• Plan and test storage-heavy AI setups: KV cache tiering, faster checkpointing, high-throughput data pipelines over RDMA and NVMeoF
• Set system-level designs across GPUs, CPUs, DPUs, NVLink interconnects and Spectrum-X Ethernet to cut waste over the AI lifecycle
• Run prototyping, benchmarking and performance validation at scale, tracing bottlenecks in compute, network and storage

What they look for

• Twelve or more years architecting datacenter-scale AI, HPC or storage infrastructure in a principal or solutions architect role
• Bachelor's in computer science or similar, or the equivalent in experience
• Hands-on with RDMA (RoCEv2 or InfiniBand), NVMeoF, GPFS, Lustre or S3-compatible object storage, plus low-latency fabric design
• Working knowledge of KV cache management and tiered memory and storage hierarchies for inference

Nice to have

• Experience with RAG pipelines and vector database integration
• Familiarity with DOCA and DPU-offloaded data services such as storage acceleration and telemetry