Senior GPU Memory Architect

Overview In this role you will advance GPU memory systems at NVIDIA, shaping next-gen architectures for AI, autonomous vehicles, gaming, and mobile platforms. You will collaborate with a distributed, product-driven team to optimize performance, power, area, and complexity across silicon processes. You’ll run performance simulations, develop high-level models, and translate findings into architectural specifications and industry-aligned interfaces. This is a hands-on role that blends research, design, and cross-functional collaboration to push the state of the art in memory subsystems. You’ll work in a dynamic environment with mentoring opportunities and a clear impact on NVIDIA’s differentiating GPUソ Compensation / Benefitscompetitive salariesgenerous benefits packageequitylocation-based compensationcareer growth opportunitiesdynamic, diverse workplace ResponsibilitiesDevelop architecture and micro-architecture for GPU memory systems focusing on performance, power, area, and schedulePerform performance simulations to improve memory access efficiencyCreate and maintain high-level functional and performance modelsAnalyze benchmarks and workloads to identify microarchitectural optimizationsDesign experiments, present results, and propose improvementsDraft architectural specifications and customer-facing documentsCollaborate with industry partners to align software interfaces with GPU memory systemsDebug performance and functional issues using high-level models, RTL simulation, and silicon Key requirements6+ years of experience in GPU or CPU architecture with memory caching and interconnectsMaster degree or equivalent in Electrical Engineering, Computer Science, Computer Engineering or related fieldPhD in computer architecture is a plusStrong communication and interpersonal skills and ability to work in a distributed, product-oriented teammentoring history is a plusstrong communicationinterpersonal skillsmentoring junior engineersmemory caching and interconnectshardware memory management unitprefetching