Custom Platform Performance Engineer

• Work with external customers to understand and characterize their unique user workloads, translating that understanding into resource utilization.
• Quickly model user workloads on existing hardware at both IP and SoC levels.
• Collaborate with cross-functional teams to propose modifications to existing IP or new IP to better meet user workloads, and provide performance improvement estimates.
• Optimize external customers’ applications, including CPU, GPU (CUDA), or machine-learning network optimizations.

• Master’s Degree in Computer Engineering or Electrical Engineering (or equivalent experience).
• 6+ years of relevant work experience focused on CPU, GPU, and HPC architectures.
• Understanding of the many factors influencing power efficiency at chip, system, and product levels.
• Strong interpersonal, communication, and collaboration skills.
• A drive to continuously learn and expand architectural breadth and depth.
• High-level understanding of computer architecture, including CPUs, GPUs, AI accelerators, fabric, DRAM, and high-speed SERDES (PCIe, chip-to-chip).
• Experience with workload analysis and characterization, and familiarity with tools and theory for workload analysis. ARM-based tools a plus but not required.
• Experience developing and using cycle-accurate and cycle-approximate performance models, and their trade-offs.

Preferred Experience

• Experience optimizing code (CPU, CUDA, ML networks) for higher performance.
• Experience defining custom accelerators, including SW/HW interface and programming model.
• Datacenter TCO calculation.
• Pre-Silicon Performance Verification.
• Experience developing plans and strategies with Architecture and Design teams to address performance issues.
• Experience creating methodologies for performance analysis.

The base salary range is $156,000 USD - $247,250 USD, determined by location, experience, and comparable positions. You will also be eligible for equity and a comprehensive benefits package.