Architect - GPU Performance
NVIDIAJob Title
Architect - GPU Performance
Role Summary
Responsible for system-level performance analysis and bottleneck identification for high-performance GPUs and SoCs. The role works with architecture and design teams to develop and validate performance models, workloads, testbenches, and tooling to optimize system performance, area, and power.
Work includes pre-silicon and silicon performance evaluation across graphics, machine learning, automotive, video, and vision workloads.
Experience Level
Mid-level — typically requires 3+ years of relevant industry experience in performance analysis or complex SoC/GPU architecture work.
Responsibilities
Primary responsibilities include analyzing system performance, building models and infrastructure, and collaborating with design teams to improve products.
- Perform system-level performance and bottleneck analysis for GPUs and SoCs across multiple workloads.
- Develop and validate hardware models at various abstraction levels (performance models, RTL testbenches, emulators, silicon).
- Create workloads and test suites targeting graphics, ML, automotive, video, and computer vision use cases.
- Collaborate with architecture and design teams to evaluate trade-offs related to performance, area, and power.
- Develop infrastructure: performance models, testbench components, analysis and visualization tools.
- Establish methodologies to improve turnaround time, select representative datasets, and enable early performance analysis.
Requirements
Key technical skills and experience expected for the role.
- Minimum ~3 years experience with performance analysis and complex SoC and/or GPU architectures.
- Strong understanding of SoC architecture, graphics pipeline, memory subsystem architecture, and NoC/interconnect design.
- Proficient in programming with C/C++ and scripting with Python or Perl.
- Strong debugging and data/statistical analysis skills; experience using RTL dumps for failure diagnosis.
- Experience in developing performance analysis infrastructure and tooling.
- Nice-to-have: exposure to Verilog/SystemVerilog and SystemC/TLM.
- Nice-to-have: experience building performance simulators or cycle-accurate/approximate models for pre-silicon analysis.
Education Requirements
BE/BTech or MS/MTech in a relevant field; PhD is a plus. Equivalent practical experience is acceptable.
About the Company
Company: NVIDIA
Headquarters: Santa Clara, California, USA
NVIDIA is a global leader in accelerated computing, renowned for its innovative solutions in AI and digital twins that transform diverse industries. The company specializes in networking technologies, providing end-to-end InfiniBand and Ethernet solutions for servers and storage that optimize performance and scalability. NVIDIA serves sectors such as high-performance computing, enterprise data centers, and cloud computing, constantly reinventing its products and services to stay ahead in the market.
