Waymo is an autonomous driving technology company with the mission to be the world's most trusted driver. Since its start as the Google Self-Driving Car Project in 2009, Waymo has focused on building the Waymo Driver—The World's Most Experienced Driver™—to improve access to mobility while saving thousands of lives now lost to traffic crashes. The Waymo Driver powers Waymo’s fully autonomous ride-hail service and can also be applied to a range of vehicle platforms and product use cases. The Waymo Driver has provided over ten million rider-only trips, enabled by its experience autonomously driving over 100 million miles on public roads and tens of billions in simulation across 15+ U.S. states.
The Perception team builds the system which learns the spatial-temporal representation and their semantic meanings of the surrounding environment of the autonomously driving vehicle (ADV), i.e., the system that “perceives” the world around the car. We work jointly with downstream teams on the optimization and integration into the Waymo Driver. We conduct our own research to address real-world problems and collaborate with research teams at Alphabet. We have access to millions of miles of driving data from a diverse set of sensors, enabling engineers like you to (1) develop methods for efficiently and continuously learning from large scale real-world data, to (2) develop models and model training at scale, to (3) analyze real-world behavior and develop systems for handling the complexities of interacting with the real-world, and (4) optimize models for our onboard and offboard hardware.
The DRAW team is a perception problem-domain team - our name stands for Degraded Road Surfaces and Weather. We take a high-level business problem like “waymo vehicles need to drive safely in snow” and use whatever technologies and tools we need to solve the problem at hand. Most of our work is ML-related. Recently we have been working with both large supervised multi-model models (lidar+camera+radar) as well as few-shot detection using Vision Language Models (VLMs). We work closely with perception platform teams that build infra for us as well as behavior teams that focus on changing the car’s behavior in response to new outputs we produce.
Previously, our main focus was on driving in rain and dense fog. The team built Waymo’s first ever ML weather estimators to determine the weather around the vehicle and set the vehicle's speed appropriately. We also developed signals to determine when our sensors are in need of cleaning. Finally, we delivered ML models that let the vehicle avoid floods and puddles. We have also worked to make the waymo ADV driver appropriately around potholes, sand, trenches, debris, etc. Our new focus area is snow - we need to build models to understand road friction, snow accumulation, tire tracks in snow, etc.
In this hybrid role you will report to a Technical Lead Manager.
You will:
-
Apply machine learning techniques to build multi-modal sensor fusion architectures and spatial-temporal representation learners for object detection and tracking, occupancy and semantic segmentation, road understanding, etc.
-
Develop scalable recipes for large data, large model training running on Alphabet’s compute infrastructure, create methods and recipes for pre-training and post-training.
-
Develop methods and recipes for distributed fine-tuning enabling multiple developers to simultaneously improve the model, develop methods and recipes to avoid regression against a production system.
-
Develop and maintain model evaluation recipes and metrics for measuring and improving performance of pre-trained and fine-tuned models
You have:
-
Bachelors in Computer Science or a similar discipline, or an equivalent amount of deep learning experience
-
5+ years experience in Machine Learning and/or Computer Vision
-
Experience with Python
-
Experience with ML frameworks like PyTorch, JAX, or Tensorflow.
We prefer:
-
MS or PhD Degree in Machine Learning, Robotics, Computer Science or a similar discipline
-
Publications at top-tier conferences like CVPR, ICCV, ECCV, ICLR, ICML, ICRA, IROS, RSS, NeurIPS, AAAI, IJCV, PAMI
-
Experience with C++
