AI Manipulation Engineer (human)

About the role

\n\nYour mission & challenges • Policy Design & Training: You design and train the learning-based policies that map multimodal sensing, from vision and language down to raw tactile and IMU data, into smooth, precise, and safe hardware actions. Multi-Contact Manipulation: You sit at the intersection of imitation learning, reinforcement learning, and physical deployment. Your problem is the hard one: multi-contact manipulation on real, underactuated, tactile-rich hardware. Dexterous Hand Control: You teach NEURA's hands to manipulate the world with human-like dexterity • Collaboration & Execution: You work closely with ML, robotics, and software teams to deliver trained policies that work on the hands. What we can look forward to • Master's or PhD in Robotics, Computer Science, Machine Learning, or a related field with a strong focus on robotic manipulation or reinforcement learning • Hands-on experience training manipulation policies with imitation learning or deep RL on physical robot arms or dexterous hands • Deep familiarity with GPU-accelerated simulation (Isaac Lab and Isaac Sim, or MuJoCo), including building custom environments and assets • Strong PyTorch, with experience using robot-learning libraries such as Stable-Baselines3, Ray RLlib, or LeRobot • Solid foundations in kinematics, dynamics, spatial transforms, and closed-loop control • Proficient Python and C++, clean reproducible code, Git, and Docker Nice to have: • Experience with teleoperation hardware such as VR controllers, data gloves, or vision-based hand tracking • Experience training or fine-tuning VLA or diffusion-based architectures • Experience with tendon-driven or highly underactuated mechanical systems