GPU Power Architect

Intel Intel · Semiconductors · California, Folsom, United States

The role focuses on designing and developing energy-efficient hardware architectures for AI/ML workloads, specifically for GPUs. Responsibilities include building and validating GPU power models, optimizing for performance-per-watt, and developing scalable power analysis flows. The position requires a strong background in computer architecture, digital logic design, and power modeling.

What you'd actually do

  1. Design and develop energy-efficient hardware architectures for AI/ML workloads.
  2. Build and validate GPU power models across representative use cases and workloads.
  3. Architect hardware mechanisms to control ICCmax and mitigate di/dt transients.
  4. Conduct detailed workload power analysis and drive optimizations to achieve performance-per-watt (perf/W) targets.
  5. Develop scalable flows and methodologies for accurate and efficient power analysis.

Skills

Required

  • Computer architecture
  • Digital logic design using SystemVerilog
  • Development and validation of GPU power models
  • Electrical and thermal constraints in datacenter-class GPUs
  • Python proficiency

Nice to have

  • Ph.D. degree
  • innovate and drive strategic advancements in GPU architecture
  • collaborating across diverse engineering teams
  • evaluating tradeoffs between performance, power, and area constraints
  • Strong communication skills

What the JD emphasized

  • energy-efficient hardware architectures for AI/ML workloads
  • GPU power models
  • performance-per-watt (perf/W) targets

Other signals

  • GPU Power Architect
  • AI/ML workloads
  • GPU power models
  • performance-per-watt (perf/W) targets