Embodied AI Robot System Intern

Intel Intel · Semiconductors · Beijing, China

Develops and integrates large models (LLMs/VLMs/VLAs) into ROS 2-based robotic systems for perception, planning, and execution. Designs reward functions, training curricula, and evaluation protocols for embodied tasks, with a focus on training RL policies for manipulation.

What you'd actually do

  1. Design and develop robotic systems based on ROS 2, covering perception, planning, and execution pipelines.
  2. Develop task and motion integration modules, including high-level planning and low-level policy execution.
  3. Integrate large models (LLMs / VLMs /VLAs) into robotic systems for decision-making, grounding, and action generation.
  4. Design reward functions, training curricula, and evaluation protocols for complex embodied tasks.
  5. Debug, profile, and optimize real-world or simulated robotic systems to improve stability, latency, and success rate.

Skills

Required

  • Python programming
  • deep learning frameworks
  • ROS or ROS 2
  • Robot perception (vision, detection, segmentation, VLMs)
  • SLAM / navigation / manipulation
  • Simulation environments (MuJoCo, Isaac Gym, or similar)
  • Training RL policies for manipulation tasks

Nice to have

  • Coordinate frames and transformations (TF)
  • Basic Kinematics, Statics, Dynamics and Motion control

What the JD emphasized

  • real project experience expected
  • Training RL policies for manipulation tasks

Other signals

  • integrating large models into robotic systems
  • design reward functions, training curricula, and evaluation protocols for complex embodied tasks
  • training RL policies for manipulation tasks