Reinforcement Learning Intern

NVIDIA is seeking a Reinforcement Learning Intern to join its mission in building intelligent, safe, and efficient self-driving technology. As an intern, you will research, prototype, and implement state-of-the-art Reinforcement Learning (RL) algorithms for autonomous vehicle decision-making and trajectory planning.

Responsibilities:

• Research, prototype, and implement state-of-the-art Reinforcement Learning (RL) algorithms for autonomous vehicle decision-making and trajectory planning.
• Assist in developing and scaling training pipelines and simulation environments for complex RL experiments.
• Collaborate closely with senior researchers and engineers across perception and planning teams to integrate RL models into the autonomous driving stack.
• Analyze and benchmark RL model performance against imitation learning and traditional baselines in complex urban simulation scenarios.
• Explore techniques to optimize and lightweight RL policies for real-time deployment on automotive hardware.

Requirements:

• Currently pursuing an MS or PhD in Computer Science, Robotics, Electrical Engineering, or a related quantitative field with a strong academic background.
• Solid theoretical foundation in Reinforcement Learning, including policy gradients (e.g., PPO, GRPO), actor-critic methods, and model-based RL.
• Hands-on project or academic research experience developing and training deep RL algorithms using PyTorch or TensorFlow.
• Strong programming skills in Python and foundational familiarity with C++.
• Great eagerness to learn, strong problem-solving abilities, and excellent communication skills for collaborative team research.

Nice to Have:

• Publication record at top-tier AI/Robotics conferences (e.g., NeurIPS, ICML, IROS, ICRA, CoRL) or relevant open-source contributions in RL or autonomous driving.
• Research exposure to generative modeling (e.g., Flow Matching, Diffusion, AR decoders) applied to trajectory prediction or world models.
• Experience addressing compounding errors, distribution shift, or safety constraints in robotic control and autonomous navigation simulation environments.