Description
NVIDIA's Performance Group is seeking a technical NIC & DPU Performance Research & Analysis Engineer to drive end-to-end performance strategy, characterization, optimization, and HW offload accelerations for next-generation network interfaces and Data Processing Units (DPUs).
In this role, you will analyze and accelerate critical networking and storage technologies and protocols powering multi-node AI clusters. You will profile performance across the complete hardware-to-software execution stack,stretching from DPU Arm CPU subsystems and PCIe to high-speed RDMA, SPDK, and DOCA acceleration pipelines.
Responsibilities
- Data-Path Optimization: Drive end-to-end benchmark plans, performance characterization, expectations, and optimizations for NICs and DPUs deployed in distributed AI and cloud data centers.
- Full-Stack Profiling: Profile low-level memory, PCIe, host CPU, and DPU ASIC execution paths to identify micro-architectural stragglers, queue bottlenecks, and tail-latency outliers.
- Storage & Network Acceleration: Hands-on acceleration and offloading of storage (SPDK, NVMe-oF, VirtIO) and networking protocols (RDMA/RoCEv2, TCP/IP, DOCA).
- AI-Assisted Workflow Automation: Leverage AI agents, autonomous scripts, and automated regression frameworks to scale continuous performance evaluation.
- Cross-Functional RCA: Partner closely with Architecture, ASIC, Firmware, and Software engineering groups to conduct rigorous Root Cause Analysis (RCA) and establish performance targets for upcoming pre- and post-silicon platforms.
Requirements
- B.Sc. or M.Sc. in Computer Science, Computer/Electrical Engineering, Software Engineering, or equivalent practical experience.
- 5+ years of experience in low-level networking, storage acceleration, and systems-level performance tuning in cloud or HPC environments.
- Direct hands-on experience with at least two of the following: DOCA, SPDK, VirtIO, RDMA/RoCEv2, TCP/IP.
- Strong programming proficiency in C, Python, and Linux shell scripting (Bash).
- Micro-architectural understanding of modern CPU/ASIC memory hierarchies, PCIe Gen 5/6 interconnects, and Linux kernel networking stacks.
- Proven track record using AI coding agents or automated script generators to streamline daily engineering workflows.
Preferred Qualifications
- Deep hands-on experience tuning NVIDIA BlueField DPUs, ConnectX SmartNICs, or custom hardware offload engines.
- Mastery of the full SPDK & NVMe-oF (NVMe over Fabrics) storage stack over high-speed RDMA transport layers.
- Comprehensive system-level understanding across x86 (Intel/AMD) and Arm architectures, PCIe topology, and GPUDirect Storage (GDS).
- Proven experience isolating stragglers and tail latency in multi-tenant, cloud-scale environments.