Software Engineer III, ML, TPU Efficiency, YouTube

YouTube builds the machine learning systems that power the YouTube algorithm, and this role focuses on making those models more efficient across training and serving. You will help reduce training and serving costs while improving fleet utilization by working at the intersection of ML model optimization and TPU compiler co-design. The position is based in Mountain View, CA (onsite).

Compensation: USD 147,000 - 210,000 per year, plus 15% bonus target, equity, and benefits.

Responsibilities

• Profile ML workloads to find compute and memory bandwidth bottlenecks, then optimize accelerator utilization to maximize compute efficiency.
• Explore, implement, and productionize algorithmic efficiency techniques such as low-precision quantization, knowledge distillation, parameter sharing, and attention optimizations.
• Optimize auxiliary serving and distributed data pipelines, including data ingestion, feature transformation, embedding lookups, and memory caching to support real-time training and inference.
• Collaborate with ML model developers and researchers to co-design hardware-friendly model architectures and deploy universal efficiency libraries.

Requirements

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

Preferred Qualifications

• Master's degree or PhD in Computer Science or related technical fields.
• Experience developing accessible technologies.
• Experience debugging correctness and performance issues across the full ML software stack.
• Experience with ML compilers and their internals, including writing compiler optimization passes.
• Familiarity with accelerator hardware architectures (TPUs/GPUs).

Tech Stack

• C++, Python
• TensorFlow, JAX, PyTorch
• XLA
• TPUs, GPUs

In most instances, this position requires in-person interviews as part of the hiring process.