Principal Machine Learning Engineer, Geometric Vision

Wayve•Sunnyvale, CA

2d•$407,330 - $460,020•Hybrid

About The Position

As a Principal Engineer on the Model Foundations team, you will build the geometric vision and 3D foundation models that underpin our autonomous driving systems. You will work at the intersection of large-scale deep learning, geometric computer vision, and real-world robotics, developing models that learn 3D structure and dynamics from fleet-scale sensor data. You will be a hands-on technical leader. You will set direction for geometric vision, prototype and train new model architectures, build the data and supervision needed to scale them, and take successful ideas through to deployment on real vehicles.

Requirements

Deep expertise in 3D computer vision, geometric vision, or 3D machine learning, with experience in areas such as multi-view geometry, neural rendering, reconstruction, implicit representations, or world modeling.
Strong experience designing, training, and evaluating modern deep-learning models at scale, using PyTorch or a comparable framework.
Strong mathematical and technical foundations in geometry, linear algebra, probability, optimization, and 3D transformations, combined with excellent software engineering skills in Python and C++
A track record of taking difficult research problems from idea to working system, including building large-scale data, training, evaluation, or deployment pipelines.
Principal-level technical leadership: the ability to identify high-leverage problems, set research and engineering direction, make strong architectural decisions, and raise the technical bar across teams.

Nice To Haves

3D and geometric vision: multi-view geometry, dense 3D reconstruction, neural fields, NeRFs, Gaussian Splatting, or feedforward 3D models.
Foundation and world models: large-scale vision pre-training, self-supervised learning, video models, generative models, or learned scene dynamics.
Geometric data engines: offline SLAM, structure-from-motion, reconstruction, calibration, auto-labeling, and large-scale ground-truth generation.
Multimodal perception: learned representations and fusion across camera, radar, LiDAR, and other sensing modalities.
Production ML systems: distributed training, large-scale experimentation, and deploying neural networks on real-time, resource-constrained hardware.

Responsibilities

Design and train 3D foundation models and world models using large-scale driving data.
Develop model architectures for 3D perception, geometric reasoning, reconstruction, and world modeling across space and time.
Build scalable data generation and auto-labeling pipelines that produce high-quality geometric supervision from large volumes of sensor data.
Develop and scale offline SLAM and 3D reconstruction systems and pipelines, using large-scale sensor data to recover accurate trajectories, scene geometry, calibration signals, and geometric supervision for model training and evaluation.
Develop and apply techniques in multi-view geometry, neural rendering, NeRFs, Gaussian Splatting, implicit 3D representations, and feedforward 3D modeling.
Explore geometry-aware tokenization and representation learning, including efficient ways to encode and fuse information across cameras, viewpoints, time, and sensing modalities.
Develop foundation vision models that make effective use of camera, radar, LiDAR, and other sensor data for learning rich representations of the physical world.
Explore video and generative modeling approaches for learning scene structure, dynamics, and future evolution from driving data.
Train and evaluate models at scale on distributed compute, rapidly iterating on architectures, objectives, data, and training recipes.
Develop automated evaluation and ground-truth systems for measuring geometric consistency, reconstruction quality, 3D understanding, and downstream driving performance.
Optimize and deploy models into production autonomous-driving systems, working across model architecture, inference, and onboard constraints.
Set technical direction for geometric vision at Wayve and work closely with researchers and engineers across foundation models, perception, simulation, data, sensing, and deployment.