Role Summary:
Engineers in this role will design, implement, and optimize advanced AI algorithms that enable humanoid robots to perceive, plan, and execute manipulation tasks using cutting-edge techniques like reinforcement learning (RL), diffusion/flow models, and vision-language-action (VLA) models. They'll bridge research and real-world execution, focusing on robust, scalable manipulation behaviors.
Key Responsibilities:
- Develop manipulation algorithms for high-DOF humanoid hands (grasping, pick/place, bimanual tasks).
- Build and train RL-based policies and generative models (diffusion/flow) for robust manipulation.
- Integrate VLA models into perception-to-action pipelines for instruction-driven manipulation.
- Create and maintain sim-to-real pipelines using simulation tools (MuJoCo, IsaacSim/Lab, etc.).
- Collaborate with control, perception, and hardware teams to deploy algorithms on real robots.
- Evaluate performance metrics, conduct experiments, and iterate based on real-world tests.
Qualifications:
- Bachelor's / Master's / PhD in Robotics, Computer Science, Mechatronics, or related field.
- 2+ years of experience in AI/robotics manipulation.
- Strong programming skills in Python and/or C++; experience with ML frameworks (PyTorch, TensorFlow).
- Experience with RL training, reward engineering, and simulation environments.
- Knowledge of ROS/ROS2 and real-robot deployment.
- Solid understanding of motion planning, kinematics, and perception systems.
Preferred:
- Prior work with VLA or foundation models in robotics.
- Publications in relevant AI/robotics conferences.
- Familiarity with real-world robotics testing and hardware integration.
