Senior Systems Performance Engineer
About The Position
NVIDIA is seeking a Senior Systems Performance Engineer for its DGX Server Product Engineering Team. The role works with hardware and software engineers to develop and implement complex automated test plans for GPU-accelerated computing products. The engineer develops GPU workloads and AI application recipes to stress targeted hardware subsystems, adapts AI workloads to new GPU platforms, and uses architectural analysis and performance models to characterize behavior and identify bottlenecks. The role also involves designing controlled experiments, correlating model predictions with workload profiles and system telemetry, investigating performance anomalies and failures, and building automation for workload deployment, experimentation, telemetry collection, and analysis.
Requirements
- Bachelor’s degree in Computer Engineering, Electrical Engineering, Computer Science, or a related field, or equivalent experience.
- 5+ years of relevant experience in systems performance engineering, performance modeling, accelerated computing, or complex system development and debug.
- Deep understanding of computer architecture, including GPU and CPU execution, memory hierarchies, interconnects, and their impact on application performance.
- Experience developing analytical or simulation-based performance models and validating predictions against measurements from running systems.
- Hands-on experience deploying, profiling, and tuning AI training or inference workloads using frameworks such as PyTorch, TensorRT-LLM, vLLM, or SGLang.
- Strong Python programming and data analysis skills, including designing controlled experiments, interpreting telemetry, and communicating findings.
- Working knowledge of CUDA and GPU profiling tools, with the ability to connect application behavior to hardware utilization and bottlenecks.
- Experience debugging workloads and system behavior in Linux environments and collaborating across hardware and software teams.
Nice To Haves
- Experience with NVIDIA Dynamo, TensorRT-LLM, vLLM, SGLang, or NCCL.
- CUDA kernel development or optimization experience using cuBLAS, CUTLASS, or related libraries.
- Familiarity with x86/Arm servers, NVLink, PCIe, InfiniBand/Ethernet, and distributed GPU systems.
- Experience using hardware error, power, thermal, and reliability telemetry to develop targeted stress coverage and isolate failures.
- A track record of building automation and turning experimental findings into measurable improvements in performance, coverage, or product quality.
Responsibilities
- Develop and tune synthetic GPU workloads using CUDA or GPU programming DSLs, and application workload recipes using AI training and inference frameworks, to stress targeted hardware subsystems and expand validation coverage.
- Onboard and adapt state-of-the-art AI training and inference workloads to new GPU platforms.
- Apply architectural analysis and develop system-level performance models to characterize workload behavior, identify bottlenecks, and guide workload selection and stress coverage.
- Design controlled experiments and correlate model predictions with workload profiles, system telemetry, and measured results; investigate differences and refine workloads, models, and test strategies.
- Analyze performance anomalies, workload correctness, and hardware and software failures; create reproducible cases and partner with architecture, design, and software teams to drive fixes and verify improvements.
- Build automation for workload deployment, experiment execution, telemetry collection, and analysis across systems and configurations.
Benefits
- Comprehensive benefits package.
- Equity.
Stand Out From the Crowd
Upload your resume and get instant feedback on how well it matches this job.
What This Job Offers
Job Type
Full-time
Career Level
Senior
