Architect - GPU Performance
NVIDIA
India, BengaluruPosted 4 months ago
Skill Required
Machine LearningDeep LearningR ProgrammingdebuggingPythonVerilogdesignC++PerlFulltime
Key highlights
- Level Required: 3+ years
- Key benefit: Work on next-gen GPUs/SoC systems impacting automotive and AI domains
- Notable requirement: Expert knowledge of SoC/GPU architecture, graphics pipeline, and NoC/Interconnect design
Role overview
NVIDIA is seeking a Systems Performance Engineer to work on high-performance GPUs and System-on-Chip (SoC) systems. The role involves analyzing system performance, identifying bottlenecks, developing performance models and testbenches, and collaborating with architecture teams to optimize CPUs, GPUs, and interconnect fabrics. Responsibilities span multiple domains including graphics, machine learning, automotive, and high-performance computing (HPC), with a focus on enabling scalable and parallel computing solutions. The ideal candidate will contribute to innovative projects that enhance next-generation visual computing and automotive systems while operating in a collaborative, diverse, and mission-driven environment.
Responsibilities
- Conduct system-level performance analysis and bottleneck identification for complex GPUs and SoCs
- Develop hardware models across abstraction levels (performance models, RTL test benches, emulators) to analyze and improve system performance
- Define and develop workloads/test suites for graphics, machine learning, automotive, video, and computer vision applications
- Collaborate with architecture and design teams to evaluate trade-offs in performance, area, and power consumption
- Build performance models, testbench components, and analysis/visualization tools
- Drive initiatives to reduce turnaround time, identify representative datasets, and enable early performance analysis in product development
Requirements
- BE/BTech or MS/MTech in a relevant field (PhD is preferred)
- 3+ years of experience in performance analysis of complex SoC/GPU architectures
- Deep understanding of SoC architecture, graphics pipeline, memory subsystem, and Network-on-Chip (NoC)/Interconnect architecture
- Proficiency in C/C++ and scripting languages (Perl/Python); experience with Verilog/SystemC/TLM is highly desirable
- Strong debugging and analytical skills, including data and statistical analysis of RTL dumps
- Experience developing performance simulators and cycle-accurate/approximate pre-silicon models (preferred)
Nice to have
- Exposure to Verilog/SystemC/TLM
- Experience with cycle-accurate/approximate performance models for pre-silicon analysis
Additional details
- NVIDIA emphasizes a culture of continuous innovation and autonomy
- Opportunity to impact next-generation visual computing, automotive, and HPC systems
- The role involves working in a diverse, supportive team environment
- Equal opportunity employer with inclusive policies (does not discriminate based on race, religion, gender, etc.)
- #LI-Hybrid indicates possible hybrid or remote work arrangement