Senior Product Architect, Storage

NVIDIA•Santa Clara, CA

9d•$224,000 - $356,500

About The Position

NVIDIA is seeking an AI Storage Platform Architect to connect its hardware platforms with real-world AI deployments. The role translates the capabilities of Rubin GPUs, Vera CPUs, BlueField DPUs, NVLink fabric, and Spectrum-X networking into validated, production-ready blueprints. Working with storage ecosystem partners, the architect will co-develop reference architectures for the NVIDIA AI Data Platform and related solutions, optimizing compute, fabric, memory, and storage for modern AI workloads.

Requirements

12+ years of experience architecting datacenter-scale AI, HPC, or storage infrastructure as a Principal Architect, Solutions Architect, Principal Engineer, or equivalent.
Bachelor’s degree in Computer Science or a related field, or equivalent experience.
Deep expertise building AI infrastructure, including disaggregated inference architectures, LLM training pipelines, and autonomous AI system patterns.
Hands-on experience with RDMA (RoCEv2/InfiniBand), high-performance storage protocols (NVMeoF, GPFS, Lustre, or S3-compatible object storage), and low-latency fabric design.
Strong understanding of KV Cache management strategies, including tiered memory/storage hierarchies for inference optimization.
Familiarity with Retrieval-Augmented Generation (RAG) architectures and the storage, indexing, and retrieval patterns they require at scale.
Experience with NVIDIA DOCA or equivalent DPU/SmartNIC programming frameworks for offloading data plane and storage services.
Proven networking foundation, including Spectrum-X Ethernet, InfiniBand, NVLink Switch fabrics, congestion control, and datacenter topologies.

Nice To Haves

Proven experience designing reference architectures jointly with storage or infrastructure OEM partners such as NetApp, DDN, VAST, Pure Storage, or Dell.
Hands-on deployment experience with disaggregated inference systems, including prefill/decode separation, KV Cache offload, and request routing.
Deep familiarity with NVIDIA Grace-Hopper, Grace-Blackwell, or upcoming Vera-Rubin platforms and their system-level implications for AI workloads.

Responsibilities

Architect end-to-end reference architectures for disaggregated inference aligned with NVIDIA Dynamo, large-scale foundation model training, and agentic AI pipelines in collaboration with storage and ecosystem partners.
Design and validate storage-optimized AI infrastructure, including KV Cache tiering, checkpoint acceleration, and high-throughput dataset pipelines using RDMA and NVMeoF fabrics.
Define system-level architectures spanning Rubin GPUs, Vera CPUs, BlueField DPUs, NVLink interconnects, and Spectrum-X Ethernet to improve efficiency throughout the AI lifecycle.
Develop and publish reference architectures, whitepapers, and deployment guides for the NVIDIA AI Data Platform and partner-integrated solutions.
Drive prototyping, benchmarking, and performance validation of AI infrastructure at scale, diagnosing bottlenecks across compute, networking, and storage.
Use DOCA to architect DPU-offloaded data services, including storage acceleration, telemetry, security enforcement, and network virtualization.
Collaborate with RAG and autonomous AI teams to build retrieval-optimized storage architectures, including vector database integration, low-latency object access patterns, and inference-aware caching.
Partner with customers and ecosystem collaborators to co-innovate and deliver proof-of-concepts and MVPs demonstrating end-to-end AI platform performance.