Architect - GPU Performance

NVIDIA is seeking an Architect - GPU Performance to join their Deep Learning Automotive team in Bengaluru. This role focuses on system-level performance analysis and bottleneck identification for complex, high-performance GPUs and System-on-Chips (SoCs). You will contribute to the development of next-generation visual computing, automotive, GPU, and HPC systems.

Responsibilities

• Perform system-level performance and bottleneck analysis on GPUs and SoCs.
• Develop and utilize hardware models at various abstraction levels, including performance models, RTL test benches, emulators, and silicon.
• Create workloads and test suites targeting graphics, machine learning, automotive, video, and computer vision applications.
• Collaborate with architecture and design teams to evaluate trade-offs in system performance, area, and power consumption.
• Build infrastructure such as performance models, testbench components, and analysis/visualization tools.
• Drive methodologies to improve turnaround time and enable early-stage performance analysis.

Required Qualifications

• BE/BTech or MS/MTech in a relevant field (PhD is a plus).
• Minimum 3 years of experience in performance analysis and complex SoC or GPU architectures.
• Strong understanding of SoC architecture, graphics pipelines, memory subsystems, and NoC/Interconnect architecture.
• Expert proficiency in C/C++ and scripting languages (Perl/Python).
• Strong debugging and data/statistical analysis skills, including experience with RTL dumps.

Preferred Qualifications

• Experience with Verilog/System Verilog and SystemC/TLM.
• Hands-on experience developing performance simulators and cycle-accurate/approximate models for pre-silicon analysis.