Back to Open Positions
Experienced Roles

Embodied AI Algorithm Engineer/Expert (Embodied Foundation Models)

Beijing
Full-time
Multimodal / Embodied AI foundation models
Relevant background

You will be responsible

  • Design and implementation of a Vision-Language-Action (VLA) model for robot decision-making and control
  • Build multimodal data and training processes to improve the model's generalization ability in real tasks
  • Promote the implementation of the model from offline training to deployment on real robots
  • Continuously optimize model performance and inference efficiency, supporting system-level integration

We hope you have

  • Background in computer science / artificial intelligence and related fields
  • Solid foundation in deep learning, with experience in multimodal or foundation models
  • Familiar with Python, familiar with PyTorch / JAX
  • Familiar with the training and use of mainstream visual language models (Qwen, etc.) or embodied models (π0, RT, etc.)

Preferred Qualifications

  • Has multimodal (vision-language-action) experience related to robots
  • Experience in large-scale data training or data engineering
  • Experience in model compression / acceleration / deployment
  • Experience with top conference papers or open source projects

Our way of working

  • Emphasize real closed-loop: algorithms should not only be effective in simulation, but also run stably on real robots.
  • Strong engineering orientation: connect the full chain from data → model → deployment, rather than just optimizing a single algorithm.
  • Rapid iteration: Move frequently between simulation and physical robots to validate and refine solutions quickly.
  • Strong collaboration: algorithms, hardware, and systems closely work together to solve complex embodied problems.

Interested in This Role?

Submit your application below and our team will review your resume shortly.