ML TPU Efficiency Engineer

Google; Mountain View (CA); On-site; USD 147,000 - 210,000; Full time

Benefits offered by this job

15% bonus target

Equity

Benefits

Job summary

Google is seeking software engineers to advance the YouTube algorithm, focusing on model efficiency and cost reduction across training and serving. You will own optimization techniques, including low-precision quantization, distillation, and parameter sharing, while collaborating with researchers to develop hardware-friendly models.

The role emphasizes building scalable ML systems and contributing to a broad stack from models to hardware/software co-design, with opportunities to influence

Qualifications

• Bachelor's degree or equivalent practical experience.
• 2 years of programming in C++ or Python.
• 2 years of software design and architecture experience.
• 2 years of testing, and launching software products.
• Experience with ML model optimization.
• Experience with ML frameworks such as TensorFlow, JAX, PyTorch or ML compilers (e.g., XLA).

Responsibilities

• Profile ML workloads, identify compute/memory bottlenecks, and optimize accelerator utilization.
• Explore and productionize efficiency techniques like quantization and distillation.
• Optimize serving and data pipelines for real-time training and inference.
• Co-design hardware-friendly model architectures and deploy efficiency libraries.

Skills

C++

Python

ML optimization

ML frameworks

Software design

Education

Bachelor's degree or equivalent

Master's degree or PhD

Tools

TensorFlow

JAX

PyTorch

XLA

Job description

Google is seeking software engineers to advance the YouTube algorithm, focusing on model efficiency and cost reduction across training and serving. You will own optimization techniques, including low-precision quantization, distillation, and parameter sharing, while collaborating with researchers to develop hardware-friendly models.

The role emphasizes building scalable ML systems and contributing to a broad stack from models to hardware/software co-design, with opportunities to influence