Senior Storage Software Engineer - DGX Cloud

NVIDIA develops accelerated computing technologies for artificial intelligence, high-performance computing, and visualization. The Senior Storage Software Engineer will serve as a hands-on technical lead contributing to open-source parallel and distributed file systems, troubleshooting storage issues across large GPU clusters, and establishing performance, configuration, and operational standards.

Responsibilities:

• Contribute to open-source file systems. Contribute code to open-source parallel and distributed file systems, and distributed object storage. Upstream fixes and features, and engage directly with the upstream communities and maintainers
• Serve as a hands-on storage software lead. Write and review production code yourself, and read kernel, NFS, NVMe-oF, or SPDK source when a bug requires it. Make the final technical calls on storage deliveries against measurable targets
• Triage and troubleshoot at scale. Triage, troubleshoot, and root-cause large, complex storage issues across very large GPU clusters (tens of thousands of GPUs) — I/O and metadata performance, data corruption, and recovery
• Validate architecture and capabilities. Validate storage architecture, capabilities, performance, and durability. Run scale tests, benchmarks, and recovery drills, and qualify new builds against measurable performance and durability targets
• Recommend configuration, tuning, and guidelines. Define and recommend configuration, tuning, and operational best practices for high-performance file systems on GPU infrastructure, and help operators and internal customers apply them
• Partner broadly. Work with training, inference, and accelerated-computing teams, site-reliability and operations, networking, and security, and collaborate with cloud providers, neocloud operators, and storage vendors on a common architecture
• Work AI-first. Use modern AI coding and agentic tools day-to-day to accelerate building, debugging, validation, and operations

Requirements:

• BS, MS, or PhD in Computer Science, Electrical Engineering, or a related field — or equivalent experience
• Over 12 years of direct experience in storage software engineering, including extensive involvement with a high-performance parallel or distributed file system handling multi-petabyte scale
• Contributions to open-source projects involving a distributed or parallel file system
• You are fully engaged in engineering tasks
• You write and review production code, examine file system, kernel, NVMe-oF, or SPDK source to identify bugs, and personally conduct scale tests or recovery drills instead of assigning them to others
• Experience diagnosing and resolving storage problems in extensive GPU or HPC clusters, including analysis of I/O and metadata performance
• Strong proficiency in at least one systems language (C, C++, Rust, or Go) and proficiency in Python; comfortable in the Linux kernel storage and networking stacks (block layer, RDMA / RoCE / InfiniBand, NVMe, page cache, VFS, multipath)
• Solid understanding of object storage (S3 / Swift-class) and block storage (NVMe-oF, iSCSI)
• Strong written and verbal communication; capable of clarifying complex technical trade-offs to engineers, SREs, vendors, and internal customers
• Comfort operating in a 24/7 production environment where storage incidents directly impact GPU availability, with a security-first approach baked into every build
• 100% hands-on engineering
• You write and review production code, read file system, kernel, NVMe-oF, or SPDK source to chase bugs, and run scale tests or recovery drills yourself rather than delegating
• Maintainers or sustained contributions to widely used public projects
• Experience crafting or operating storage for AI training or inference at very large GPU scale, with measurable gains in GPU utilization or reductions in I/O bottlenecks
• Kernel and file system development experience, metadata scalability, data placement, failure recovery, or HSM or equivalent experience
• Kubernetes and CSI driver development for storage
• Hands-on experience with SPDK, libfabric, or FUSE performance optimization