Staff ML Performance Engineer - GPU Optimization

Google LLC; Sunnyvale (CA); On-site; USD 186,000 - 228,000; Full time

Job summary

Google is seeking a Staff Software Engineer for ML Performance on GPU to drive high-performance ML workloads and next-gen GPU architectures. You will work across model deployment, evaluation, and data processing, ensuring peak efficiency on large-scale systems.

The role requires hands-on CUDA/Triton/CUTLASS experience, deep knowledge of ML workloads, and collaboration with cross-functional teams to push performance forward.

Qualifications

• Bachelor’s degree or equivalent practical experience.
• 8 years of software development experience.
• 5 years of ML design/infrastructure experience.
• GPU architectures and performance bottlenecks expertise.
• Low-level GPU programming (CUDA, Triton, CUTLASS) experience.
• Experience with deployment of LLMs on AI accelerators.

Responsibilities

• Identify and maintain LLM training and serving benchmarks; use them to identify performance opportunities and guide optimizations across GPUs.
• Partner with product teams to onboard, optimize, and scale ML models on GPU hardware.
• Conduct architecture-level simulations and roofline analyses to guide system designs.
• Analyze fleet-wide performance metrics to drive scalable optimizations across Google's infrastructure.
• Research and implement model/data efficiency techniques and profiling mechanisms.

Skills

ML design infra

GPU architectures

Low-level GPU programming

LLM deployment

Performance engineering

Education

Bachelor’s degree or equivalent practical experience

Master’s or PhD in Engineering/CS or related field

Tools

CUDA

Triton

CUTLASS

OpenXLA

XLA

Roofline analysis tools

Job description

Google is seeking a Staff Software Engineer for ML Performance on GPU to drive high-performance ML workloads and next-gen GPU architectures. You will work across model deployment, evaluation, and data processing, ensuring peak efficiency on large-scale systems.

The role requires hands-on CUDA/Triton/CUTLASS experience, deep knowledge of ML workloads, and collaboration with cross-functional teams to push performance forward.