ML Research Scientist – Prediction & Smart Agents

Nuro; California (MO); On-site; USD 194,000 - 291,000; Full time

Benefits offered by this job

Annual bonus

Equity

Benefits package

Job summary

Nuro is seeking a researcher to advance state-of-the-art models predicting surrounding traffic behavior. You will join the Prediction and Smart Agents team, building scalable models deployed onboard as part of the planning stack and used in realistic closed-loop simulation.

You will explore generative sequence modeling, multi-modal inputs, and controllable agents, collaborating with Planning to train via reinforcement learning and push dependable deployments to real robots.

Qualifications

• MSc/PhD in Computer Science, Artificial Intelligence, Mathematics, or a closely related field.
• Experience and research in sequential decision-making, prediction, imitation learning, RL, or robotics.
• Strong Python (required) and ML frameworks; C++ is a plus; PyTorch experience preferred.
• 2+ years deploying machine learning systems onboard, ideally for prediction.
• Publications in top conferences (NeurIPS, ICML, CVPR, etc.).
• Nice-to-have: Embodied AI for robotics, causal reasoning, diffusions, interpretability.

Responsibilities

• Design scalable ML-based prediction systems to generate multi-modal trajectories.
• Conduct cutting-edge research in generative sequence modeling and sequential decision-making.
• Collaborate with the Planning team to design realistic, controllable agents for closed-loop simulation.
• Enable agent training via reinforcement learning and mitigate uncertainties across components.
• Work with multiple autonomy teams to develop holistic solutions and push to real deployments.

Skills

Python

C++

PyTorch

Education

MSc or PhD in CS/AI/Math

Job description

Nuro is seeking a researcher to advance state-of-the-art models predicting surrounding traffic behavior. You will join the Prediction and Smart Agents team, building scalable models deployed onboard as part of the planning stack and used in realistic closed-loop simulation.

You will explore generative sequence modeling, multi-modal inputs, and controllable agents, collaborating with Planning to train via reinforcement learning and push dependable deployments to real robots.