Responsibilities
- Build and train machine learning models for interpreting driving scenes, including detecting objects, predicting lanes and roads, classifying 3D space, forecasting occupancy, and understanding maps in bird’s-eye-view format.
- Write robust, production-grade machine learning code to enable training, testing, and real-time inference within perception systems.
- Evaluate model behavior, detect failure patterns, and suggest enhancements to improve reliability across varied driving conditions.
- Detect and interpret dynamic elements such as vehicles, pedestrians, lanes, obstacles, and weather-related factors in driving environments.
- Use data analysis methods to assess model outputs, study data characteristics, and uncover rare or challenging scenarios.
- Support development of perception systems that fuse inputs from cameras, LiDAR, radar, and map data into cohesive environmental models.
- Utilize large datasets from simulations, vehicle fleets, and onboard sensors to refine training sets and boost model accuracy.
- Partner with data, deployment, and infrastructure teams to validate perception models under real-world driving conditions.
- Assist in embedding perception models into autonomous driving pipelines and testing frameworks to accelerate development cycles.
- Develop tools and systems that enhance training speed, experiment management, and reproducibility of results.
- Engage in technical conversations about model design, sensor fusion techniques, and training methodologies within the team.
