About The Role
You will build and evaluate humanoid robot capabilities across perception, planning, and control. This role focuses on end-to-end robot behavior development: data collection and curation, policy learning loops, safety constraints, and structured QA evaluation in simulation and on hardware to close the sim-to-real gap.
What You Will Do
-
Develop humanoid locomotion and manipulation pipelines spanning simulation, training, and deployment
-
Implement and tune whole-body control, motion planning, and contact-aware behaviors
-
Design data collection protocols and labeling schemas for robot perception and behavior learning
-
Build offline and online evaluation suites: success metrics, failure taxonomy, and regression tests
-
Improve model performance using reinforcement learning, imitation learning, and behavior cloning
-
Integrate perception outputs (vision, depth, IMU, joint states) with sensor fusion and state estimation
-
Collaborate with hardware and embedded teams to debug real-world failures and improve robustness
Required Qualifications
-
Mid-senior experience in humanoid, legged, or manipulation-focused robotics
-
Strong programming skills in Python and C++ with production-grade software engineering practices
-
Experience with ROS2/ROS, robot kinematics/dynamics, and control fundamentals
-
Familiarity with reinforcement learning, imitation learning, or policy optimization for robotics
-
Experience designing evaluation metrics, QA checks, and repeatable experiment pipelines
Preferred Qualifications
-
Hands-on experience with humanoid platforms, bipedal locomotion, or whole-body motion generation
-
Experience with simulation engines (e.g., MuJoCo, Isaac, Gazebo) and sim-to-real workflows
-
Experience with computer vision pipelines for robot perception and real-time inference
-
Knowledge of safety constraints, fall detection, collision checking, and recovery behaviors
Workplace
Remote (US). Full-time role with mid-senior expectations.