Embodied AI Research Engineer

The Role We are seeking a hands-on Embodied AI Research Engineer to develop and integrate intelligent robotic systems in simulation and on physical robots. This is an individual-contributor role for an experienced robotics engineer with broad knowledge of the robotics stack and particular strength in planning, system integration, and sim-to-real deployment. You will work on world models, robot memory, planning, locomotion, manipulation, and multimodal robot intelligence. Prior expertise in every one of these areas is not expected. The ability to understand complete robotic systems, implement new research ideas, and make them work on physical robots is more important. You will work across platforms including the Unitree G1 humanoid and robotic arms. Responsibilities Develop task-planning, hierarchical-planning, and motion-planning systems for complex robotic tasks. Integrate perception, planning, robot interfaces, and task execution into complete robotic systems. Build simulation, testing, and evaluation environments using MuJoCo, Isaac Sim, Isaac Lab, or similar. Transfer systems between simulation environments and from simulation to physical robots. Integrate and test software humanoid robots or robotic arms, and other physical robot platforms. Research and prototype world models, robot-memory architectures, Reproduce and evaluate promising methods from recent robotics and embodied-AI research. Debug issues across software, simulation, and hardware. Build maintainable robotics software, internal benchmarks, and automated experiment infrastructure. Use AI coding agents and agentic development workflows to accelerate research and engineering. Required Qualifications At least three years of relevant professional / industry experience in robotics, robot learning, reinforcement learning, control, embodied AI, or autonomous systems Practical experience with sim-to-sim and sim-to-real transfer. Experience with MuJoCo, Isaac Sim, Isaac Lab, or another robotics simulator. Experience in robotic planning, learned dynamics, sequential decision-making. hierarchical control Experience with ROS 2. Strong Python skills and practical C++ experience. Ability to translate research papers and ideas into working implementations. Strong software engineering, experimentation, testing, and debugging skills. Willingness to work directly with physical robots and troubleshoot across software, simulation, and hardware. Familiarity with AI-assisted coding tools and agentic development workflows. Preferred Qualifications Experience with quadrupeds, humanoids, robotic arms, or other autonomous systems. Experience with world models, robot memory, predictive models, or model-based reinforcement learning. Experience with motion-planning or control libraries. Evidence of strong applied work through robot demonstrations, open-source projects, publications, patents, or deployed systems. Additional useful experience includes VLA training, JEPA-style models, sensor fusion, whole-body control, dexterous manipulation, domain randomization, and large-scale training infrastructure. What We Value We are looking for someone who is curious, experimental, and comfortable challenging assumptions. You should be able to move between reading research, writing software, running simulations, and testing on physical robots without becoming attached to one technical approach before the evidence supports it. Success will be measured primarily by successful deployment on physical robots, performance against internal benchmarks, research iteration speed, and demonstrated improvements in robot autonomy, planning, locomotion, or manipulation.