Waymo
Machine Learning Engineer, GPU Kernel and Runtime
Mountain View, CaliforniaFrom $263kfull-timemidAdded today
About this role
Waymo is seeking a Machine Learning Engineer to optimize GPU kernel and runtime performance for autonomous driving systems. You'll work on low-level computational infrastructure that powers the Waymo Driver's perception and decision-making pipelines.
What you'll do
- Design and optimize GPU kernels for machine learning inference and training workloads
- Develop runtime systems for efficient execution of neural networks on automotive hardware
- Profile and benchmark GPU performance across autonomous driving use cases
- Collaborate with ML and systems teams to identify performance bottlenecks
- Implement and test kernel optimizations for production autonomous vehicles
- Contribute to hardware-software co-optimization efforts
What they're looking for
- CUDA or GPU programming (C++/Python)
- Machine learning frameworks (TensorFlow, PyTorch)
- Performance profiling and optimization
- Computer architecture and GPU design
- System-level programming and debugging
- Automotive or real-time systems experience
- Low-level optimization and assembly
- Parallel computing
Benefits
- Discretionary annual bonus program
- Equity incentive plan
- Comprehensive company benefits
- Work on cutting-edge autonomous driving technology
- Mountain View headquarters location with remote flexibility
- Collaborative environment with leading ML and systems engineers
Opens the official application on the employer’s site. No login required.
Waymo
Waymo develops autonomous driving technology and vehicles, building the AI systems, simulation platforms, and infrastructure that power the Waymo Driver. The company is hiring for ML infrastructure engineers, platform engineers, labeling system developers, backend software engineers, and automotive systems engineers to scale its autonomous driving capabilities.
- Website
- waymo.com
Likely interview questions
- Describe a time you optimized GPU kernels for latency-critical workloads—what techniques did you use and what was the performance improvement?
- How do you approach profiling and identifying bottlenecks in ML inference pipelines?