Engineer, Embodied AI

We are looking for an Embodied AI / Robot Intelligence R&D Engineer to develop next-generation robot intelligence systems that bridge large AI models with real-world robotic perception, planning, decision-making, and control. The role will focus on building a modular robot brain architecture combining LLM/VLM/VLA, world models, task planning, spatial intelligence, robot skills, physical verification, and ROS 2-based execution.

Key Responsibilities

Develop embodied AI algorithms for robotic perception, reasoning, task planning and autonomous execution.

Research and integrate LLM, VLM and VLA models into real-world robotic systems.

Develop hierarchical robot planning architectures, including LLM-based planning, task planning, skill planning and execution.

Build and maintain robot skill/capability models, task graphs and reusable skill libraries.

Research world models, spatial intelligence and persistent robot memory for long-horizon robotic tasks.

Develop semantic-geometric representations combining visual semantics, 3D geometry and robot/environment states.

Conduct simulation and real-robot experiments on mobile robots, quadruped robots, manipulators and other robotic platforms.

Track and evaluate state-of-the-art research in embodied AI, robot foundation models, VLA, world models and autonomous robotics.

Required Qualifications

Master’s degree or above in Robotics, Artificial Intelligence, Computer Science, Automation, Electrical Engineering, Mechanical Engineering, or a related field.

Strong programming skills in Python and/or C++.

Solid understanding of robotics fundamentals, including at least several of the following: robot perception; task planning; SLAM/navigation; robot manipulation; reinforcement learning.

Experience with ROS/ ROS 2.

Good understanding of modern deep learning architectures such as Transformer, VLM, LLM or generative models.

Ability to independently implement and evaluate research ideas on real robotic systems.

Experience with real-world robot deployment, publications, patents, or open-source robotics projects will be an advantage.