Join NVIDIA, a company at the forefront of visual computing, high-performance computing, and artificial intelligence. We are seeking talented engineers to contribute to our HW architecture team. You'll work on next-generation visual computing, automotive, GPU, and HPC systems, focusing on high-performance CPU and memory subsystems, advanced GPUs, and NOC-based interconnect fabrics. Make a significant impact on the future of computing and amplify human inventiveness.

As a GPU Performance Architect, you will conduct system-level performance and bottleneck analysis for complex GPUs and SoCs. You will work with hardware models across various abstraction levels, including performance models, RTL testbenches, and emulators. Your role involves understanding key product use-cases, developing targeted workloads for graphics, machine learning, automotive, video, and compute vision applications. Collaboration with architecture and design teams to balance performance, area, and power is crucial. You will also develop essential infrastructure, including performance models and analysis tools, and drive methodologies for faster turnaround times and early performance analysis.

We are looking for candidates with a BE/BTech or MS/MTech in a relevant field, with a PhD being a plus, or equivalent experience. A minimum of 3 years of experience in performance analysis and complex SoC/GPU architectures is required. Essential knowledge includes SoC architecture, graphics pipeline, memory subsystem architecture, and NoC/Interconnect architecture. Proficiency in C/C++ programming and scripting languages like Python is mandatory. Experience with Verilog/System Verilog and SystemC/TLM is highly advantageous. Strong debugging, data, and statistical analysis skills are vital for identifying and resolving issues. Developing performance simulators and cycle-accurate models for pre-silicon analysis is a significant plus.