Staff Software Engineer, TPU, Performance

Responsibilities

• Identify and maintain ML training and serving benchmarks that are representative to Google production and broader ML industry.
• Achieve performance for customer launches, and in case of third-party/Open-Source Software (3P/OSS) models, for engaged benchmark submissions ML commons, InferenceMAX, etc.).
• Use the benchmarks to identify performance opportunities and drive out-of-the-box performance toward improving the compiler, runtime, etc., in collaboration with those teams.
• Engage with Google Product teams and researchers to solve their performance problems (e.g., onboard new ML models and products on Google new TPU hardware, enabling larger models (giant models) to train efficiently on a very large-scale (i.e., thousands of TPUs).
• Analyze performance and efficiency metrics to identify bottlenecks, design, and implement solutions at Google fleet-wide scale.
• - Identify and maintain ML training and serving benchmarks that are representative to Google production and broader ML industry. - Achieve performance for customer launches, and in case of third-party/Open-Source Software (3P/OSS) models, for engaged benchmark submissions ML commons, InferenceMAX, etc.). - Use the benchmarks to identify performance opportunities and drive out-of-the-box performance toward improving the compiler, runtime, etc., in collaboration with those teams. - Engage with Google Product teams and researchers to solve their performance problems (e.g., onboard new ML models and products on Google new TPU hardware, enabling larger models (giant models) to train efficiently on a very large-scale (i.e., thousands of TPUs). - Analyze performance and efficiency metrics to identify bottlenecks, design, and implement solutions at Google fleet-wide scale.

Minimum qualifications: