About the RoleWe are building a next-generation humanoid robot platform with high-bandwidth torque-controlled joints and full-body actuation.
As a Robotics Algorithm Engineer focused on
Humanoid Whole-Body Control, you will work across
VLA / WAM, vision-based RL, whole-body control, simulation, state estimation, and real-robot deployment. You will develop learning-based systems that coordinate locomotion, manipulation, perception, and full-body motion.
We are looking for engineers with strong implementation skills, solid robotics fundamentals, and the ability to turn research ideas into reliable real-world robot behaviors.
ResponsibilitiesWhole-Body Learning & Control- Develop and deploy learning-based whole-body control policies for humanoid robots
- Coordinate locomotion, balance, torso, arms, and end-effectors
- Integrate learned policies with WBC, inverse dynamics, IK, and optimization-based control
- Develop robust contact-aware behaviors for locomotion, manipulation, and interaction
- Analyze and debug instability, contact failures, coordination issues, and policy failures
VLA / WAM & Generalist Policies- Develop and integrate VLA / WAM models for humanoid control
- Adapt foundation-model-based policies to humanoid embodiment and full-body action spaces
- Connect high-level semantic reasoning with low-level whole-body control
- Design action spaces, observations, policy interfaces, and skill representations
- Explore imitation learning, behavior cloning, diffusion policies, transformers, and RL
- Use teleoperation, demonstration, and robot interaction data for training and fine-tuning
Vision-Based Reinforcement Learning- Develop vision-based RL policies using RGB, depth, proprioception, and onboard sensing
- Build visuomotor policies for locomotion, navigation, mobile manipulation, and whole-body tasks
- Develop visual-proprioceptive representation learning and sensor fusion
- Use privileged learning, teacher-student training, distillation, domain randomization, and sim-to-real
- Improve robustness to environment variation, object variation, appearance changes, occlusion, and sensor noise
Modeling, State Estimation & Control- Apply rigid-body dynamics, contact dynamics, and humanoid kinematics to whole-body control
- Develop and integrate state estimation using IMU, encoders, force/contact sensing, and vision
- Work with floating-base dynamics and multi-contact estimation
- Combine learning-based policies with feedback control and model-based methods
Simulation, Data & Training- Build humanoid simulation and training environments using MuJoCo, Isaac Sim / Isaac Lab, or similar platforms
- Develop scalable RL, imitation learning, and visuomotor training pipelines
- Design tasks, curricula, rewards, domain randomization, and system identification
- Generate and use simulation, teleoperation, demonstration, and real-robot datasets
- Analyze sim-to-real gaps in dynamics, contact, sensing, perception, and actuators
Real Robot Deployment- Deploy whole-body and visuomotor policies on humanoid hardware with high-bandwidth torque control
- Perform real-robot tuning, debugging, system identification, and optimization
- Diagnose failures across perception, policy inference, estimation, dynamics, latency, and low-level control
- Optimize policy inference and control pipelines for real-time execution
- Work closely with perception, firmware, motor control, systems, and hardware teams
QualificationsMust Have- 3+ years of experience in robotics, controls, reinforcement learning, imitation learning, or related fields
- Strong C++ and Python skills
- Experience developing learning-based robot control policies
- Experience deploying algorithms on real robots
- Experience with whole-body control, humanoid robotics, legged robotics, or mobile manipulation
- Hands-on experience with reinforcement learning and/or imitation learning
- Solid understanding of rigid-body dynamics, floating-base systems, contact dynamics, and feedback control
- Experience with MuJoCo, Isaac Sim / Isaac Lab, or similar simulation platforms
- Familiarity with state estimation and multimodal sensing
- Strong simulation, algorithm, and hardware debugging skills
- Experience working in Linux environments
Strongly Preferred- Experience with VLA, WAM, whole-body action models, or generalist robot policies
- Experience with vision-based RL or visuomotor learning
- Experience with transformer-based policies, diffusion policies, behavior cloning, or large-scale imitation learning
- Experience combining RGB / RGB-D observations with proprioception
- Experience with humanoid locomotion and manipulation
- Experience with domain randomization, privileged learning, distillation, and system identification
- Experience with teleoperation and demonstration-data pipelines
- Experience with real-time policy deployment and GPU inference
- Experience integrating learned policies with WBC, MPC, inverse dynamics, IK, or trajectory optimization
Nice to Have- Publications or strong project experience in humanoid robotics, robot learning, RL, imitation learning, VLA, or visuomotor control
- Experience with large-scale robot datasets and multi-task policy training
- Experience with dexterous or bimanual manipulation
- Familiarity with foundation models for robotics and embodied AI
- Experience with object detection, tracking, 3D perception, or scene representations
- Experience building production-quality robotics software and deployment infrastructure
Pay Range: $80,000- $120,000 per year. The actual base salary offered will depend on factors such as the candidate's experience, skills, qualifications, and job-related considerations. This position may also be eligible for additional compensation and benefits.
