ML Research Scientist, Prediction & Smart Agents

Nuro•Mountain View, CA

16h•$193,930 - $291,150

About The Position

The mandate of the prediction team is to use advanced machine learning techniques to improve the behavior of the Nuro Driver. As a key member of the Prediction and Smart Agents team, you will focus on building state-of-the-art models for predicting the behavior of surrounding traffic. These models are crucial for our autonomous system, as they will be deployed onboard as part of our planning stack and used offboard for realistic closed-loop simulation. You will explore novel machine learning methods to solve challenging real-world problems in autonomous driving. This work includes using generative sequence modeling approaches for robustly predicting complex, interactive traffic situations. It requires deep reasoning about the intentions of other road users and how their behaviors influence safe and correct driving decisions. You will also use different input modalities, including End-to-End (E2E) approaches, for predicting other agents. A vital component of this role is building smart, controllable agents to enable effective closed-loop training in simulation. If you are passionate about solving challenging new problems, leading impactful research, and seeing your work deployed onto real robots, we encourage you to apply!

Requirements

Education: You have an M.Sc. or Ph.D. (preferable) focusing on one or more of the following areas: Computer Science, Artificial Intelligence, Mathematics, or a closely related field
Expertise: Subject matter expertise and research experience in one or more of the following: sequential decision-making, prediction, Imitation Learning, Deep Reinforcement Learning, generative modeling, large models (pretraining/finetuning), or machine learning for robotics.
Technical Skills: You have strong problem solving and programming skills in Python (required) and C++ (beneficial) and ML frameworks such as PyTorch.
Collaboration: Strong culture fit and good team player.
Experience: You have 2+ years of deploying machine learning systems onboard, ideally in the area of prediction.
Publications: Demonstrated research publications in top conferences (e.g. NeurIPS, ICLR, ICML, CVPR, RSS, CoRL, ICRA, IROS etc.)

Nice To Haves

Deep background in Embodied AI for robotics
Causal reasoning
Model interpretability and explainability
Joint prediction and planning
Understanding of Diffusion Models.

Responsibilities

Design and build scalable, machine learning-based prediction systems to generate multi-modal, realistic, and kinematically feasible trajectories.
Conduct cutting-edge research in generative sequence modeling and sequential decision-making.
Collaborate closely with the Planning team to design realistic and controllable agents for closed-loop simulation, enabling agent training via Reinforcement Learning (RL).
Mitigate accumulated uncertainties across interconnected autonomy components.
Collaborate across various autonomy teams to develop holistic solutions for top challenges, proposing ideas, prioritizing.
Derive practical, deployable solutions and see them deployed on real-world vehicles