2027 Summer Intern, PhD, Perception/Road Understanding, ML Engineer
About this role
Advanced autonomous vehicle systems require precise road understanding to safely navigate complex urban environments, a challenge this PhD internship directly addresses. Waymo seeks an incoming engineer to build and refine machine learning models dedicated to entity-centric lane geometry detection and relational topology decoding. Working within the software engineering organization, the selected candidate will tackle intricate problems in robotics, perception, and deep learning alongside industry leaders.
This role exists to advance the perception capabilities of the Waymo Driver by processing sensor data and map priors to handle real-world driving hurdles like construction zones. Ideal candidates are doctoral students in quantitative fields who possess deep technical curiosity and practical experience with neural network architectures. Those who thrive in collaborative research environments and enjoy pushing the boundaries of spatial reasoning will find a strong alignment here.
What you will do
- Design and build machine learning models by writing clean, high-performance code for entity-centric lane geometry detection and relational topology decoding.
- Set up data pipelines and train neural networks across vehicle sensor modalities and map priors using techniques like proxy auto-encoding and prior-dropout.
- Handle real-world driving challenges such as construction zones and occlusions through targeted model experimentation.
- Create structured evaluation metrics to benchmark model accuracy and topological correctness across complex intersections.
- Analyze failure cases and iterate on architectural designs to improve performance.
- Partner closely with research mentors, assigned buddies, and upstream or downstream engineering teams to evaluate downstream planning impact.
What you need
- Must be currently pursuing a PhD in Computer Science, Robotics, Electrical Engineering, Machine Learning, or a related quantitative discipline.
- Demonstrate strong programming proficiency in Python along with solid experience using modern deep learning frameworks.
- Bring hands-on experience in designing, training, and debugging deep learning architectures for Computer Vision, 3D Perception, or Graph Neural Networks.
- Possess solid foundational knowledge of 2D and 3D geometry, coordinate transformations, and spatial or relational reasoning.
Nice to have
- Track record of publications in top-tier conferences in machine learning, computer vision, or robotics such as CVPR, ICCV, ECCV, NeurIPS, ICLR, ICML, ICRA, CoRL, or AAAI.
- Experience with vectorized HD map learning, lane topology estimation, or dynamic roadgraph modeling like MapTR, TopoNet, or LaneGAP.
- Experience with large-scale distributed model training and data infrastructure including TPU or GPU clusters, Ray, Jax or Flax, and multi-GPU pipelines.
- Familiarity with autonomous vehicle perception stacks, sensor fusion involving camera and LiDAR, and downstream motion planning constraints.
Benefits
- Eligible to participate in the Company's generous benefits programs, subject to eligibility requirements.
Skills and tech stack
Questions candidates ask
What is the work location policy for this internship?
This will be a hybrid onsite internship position based in Mountain View, California.
What is the compensation for this role?
The expected hourly rate for this full-time position is $85 USD.
How does the application process work?
Resumes are accepted on a rolling basis until the role is filled. Candidates interested in multiple roles must apply to each one individually, up to their top three choices.