We are seeking an AI Research Scientist Intern (PhD) to join us in advancing the frontier of Embodied AI for robotics. This role is centered on developing next-generation robot intelligence, with a particular focus on world models, Vision-Language-Action (VLA) models, post-training, and reinforcement learning.

You will work alongside a team of world-class researchers and engineers on ambitious, real-world problems at the intersection of foundation models, decision-making, and robotics. This is an opportunity to help shape core research directions, build cutting-edge systems, and contribute to work with strong potential for publication at top-tier conferences.

Responsibilities

• Conduct research and develop advanced Embodied AI methods for robotic perception, reasoning, and control, with emphasis on:
  • World Models for action-conditioned prediction, planning, and long-horizon decision-making
  • Vision-Language-Action (VLA) models for general-purpose robotic manipulation
  • Post-training methods such as supervised fine-tuning, preference optimization, policy improvement, and online/offline adaptation
  • Reinforcement Learning for improving robustness, generalization, and task performance
• Design and execute large-scale experiments to advance robot learning capabilities across challenging manipulation and embodied reasoning tasks.
• Collaborate closely with robotics, hardware, and infrastructure teams to bring research ideas into real robotic systems.
• Evaluate new methods on real-world and benchmark tasks, and help define rigorous research standards for the team.
• Contribute to technical reports, open research discussions, and publications at leading conferences where appropriate.

Qualifications

• Currently pursuing or recently completed a PhD in Computer Science, Robotics, Machine Learning, or a related field.
• Strong research background in Embodied AI, robot learning, foundation models, or a closely related area.
• Hands-on experience with one or more of the following:
  • World models
  • Vision-Language-Action (VLA) models
  • Post-training / policy fine-tuning
  • Reinforcement learning, including offline RL, online RL, or RL for control
• Strong understanding of modern machine learning architectures, including transformers, diffusion models, and multimodal learning systems.
• Proficiency with deep learning frameworks such as PyTorch, JAX, or TensorFlow.
• Strong experimental and problem-solving skills, with the ability to independently drive research ideas from concept to evaluation.
• Requires 5 days/week in-office collaboration with the team.

Preferred Skills

• Experience with robotic manipulation, imitation learning, or large-scale robot policy learning.
• Familiarity with post-training pipelines for large models, including supervised fine-tuning, preference optimization, or RL-based policy improvement.
• Experience with simulation-to-real transfer, long-horizon decision-making, or action-conditioned prediction.
• Exposure to multimodal learning across vision, language, action, and proprioception.
• Background in bimanual manipulation, synthetic data generation, or 3D/spatial reasoning for robotics.

Why Join Us

• Work on challenging, high-impact problems at the frontier of AI and robotics
• Collaborate with a highly technical, research-driven team
• Gain hands-on experience building real systems for Embodied AI
• Opportunity to contribute to research publications and help shape emerging directions in the field
• See your work influence both long-term research and practical robotic capabilities