TPU Performance Engineer — ML Efficiency & Scale

Google; Sunnyvale (CA); On-site; USD 207,000 - 300,000; Full time

Job summary

Google is seeking software engineers for the AI and Infrastructure teams to push performance across large-scale systems and accelerator hardware. You will work across JAX and PyTorch to optimize production and research workloads, including Gemini models, across TPU fleets.

Responsibilities include designing and delivering high-performance software, collaborating with cross-functional teams, and driving scale, efficiency, and reliability in a fast-paced environment with opportunities to switch

Qualifications

• Bachelor's degree or equivalent practical experience.
• 8 years of experience in software development.
• 5 years of experience testing and launching software products.
• 5 years of experience with performance, large-scale systems data analysis, visualization tools, or debugging.
• 3 years of experience with software design and architecture.
• Experience with ML performance analysis and benchmarking.

Responsibilities

• Focus on TPU fleet efficiency analysis and performance optimization, while identifying and maintaining ML training and serving benchmarks.
• Use benchmarks to identify performance opportunities and drive out-of-the-box performance by improving the compiler, runtime, etc.
• Collaborate with Google product teams and researchers to solve performance problems and onboard new ML models onto TPU hardware.
• Analyze performance and efficiency metrics to identify bottlenecks and implement solutions at fleet-wide scale.
• Explore model and data efficiency techniques such as model co-design, quantization, and sparsity.

Skills

Software development

Testing and launching software

Performance analysis

Large-scale data analysis

Visualization tools

Debugging

Software design

Software architecture

ML performance benchmarking

Education

Bachelor's degree or equivalent practical experience

Tools

OpenXLA

Job description

Google is seeking software engineers for the AI and Infrastructure teams to push performance across large-scale systems and accelerator hardware. You will work across JAX and PyTorch to optimize production and research workloads, including Gemini models, across TPU fleets.

Responsibilities include designing and delivering high-performance software, collaborating with cross-functional teams, and driving scale, efficiency, and reliability in a fast-paced environment with opportunities to switch