Senior Deep Learning Systems Architect, Blackwell GPU Superclusters

About NVIDIA

NVIDIA is the pioneer of GPU-accelerated computing. We specialize in products and platforms for the large, growing markets of gaming, professional visualization, data center, and automotive.

Role Overview

We are looking for a Senior Deep Learning Systems Architect to join the DGX SuperPOD architecture group. In this role, you will analyze, profile, and optimize the world's most powerful AI superclusters powered by the NVIDIA Blackwell architecture, NVLink 5, and Quantum-X800 InfiniBand fabrics.

Key Responsibilities

• Profile distributed foundation model training workloads (Megatron-LM, TensorRT-LLM, DeepSpeed) across multi-thousand GPU clusters to detect architectural bottlenecks.
• Optimize Triton and CUDA kernels for mixed-precision computation (FP4, FP8, BF16) on Blackwell Tensor Cores.
• Design topology-aware communication algorithms that leverage NVLink 5 switches to minimize all-to-all communication latency.
• Collaborate with hardware engineering teams on thermal management, power profiling, and PCIe Gen6 host connectivity.
• Publish reference architectures and benchmark suites for hyperscale cloud service providers.

Qualifications & Experience

• BS, MS, or PhD in Computer Engineering, Computer Science, Electrical Engineering, or related field.
• 5+ years of experience in high-performance computing, deep learning systems, or GPU architecture.
• Proficiency in CUDA C/C++, Python, and performance profilers (Nsight Compute, Nsight Systems).
• Deep understanding of parallel algorithms (data, tensor, pipeline, sequence, and expert parallelism).

Compensation & Benefits

• Base Salary: $220,000 – $340,000 per year + valuable NVIDIA Equity (RSU) program.
• Employee Stock Purchase Plan (ESPP) with a 15% discount.
• Comprehensive healthcare coverage with health savings account (HSA) contribution.
• On-campus fitness centers, wellness days, and 401(k) matching.