Apptronik
Senior / Staff Sim-to-Real Engineer (Tendon-based Robotic Hands)
Austin, TX · Staff · Full-time · Posted Oct 30, 2026
- Experience
- Senior: 5+ years with a BS, 3+ years with an MS, or 1-2 years with a PhD. Staff: 10+ years with a BS, 8+ years with an MS, or 6+ years with a PhD.
- Education
- Senior: BS, MS, or PhD in Robotics, Electrical Engineering, Mechanical Engineering, Computer Science, or a related field. Staff: MS or PhD in those fields or a related field.
About the job
Builds and validates simulation models for tendon-driven robotic hands so reinforcement-learning control policies transfer reliably to physical hardware.
What you'll do
- Develop efficient, accurate tendon-mechanism abstractions and validate them through empirical testing and system identification.
- Characterize actuators and calibrate hand systems, including kinematics, inertial parameters, and effort production.
- Own the sim-to-real pipeline and lead technical work to improve transfer of control policies to hardware.
What they ask for
- Proven transfer of RL-based control policies from simulation to physical hardware.
- Hands-on experience with control, system identification, and characterization of Quasi-Direct Drive (QDD) actuators.
- Strong foundation in rigid multi-body dynamics, kinematic calibration, and inertial identification.
- Experience verifying effort production, characterizing friction/backlash, and diagnosing hardware-software discrepancies.
Nice to have
- Experience modeling, calibrating, or controlling tendon- or cable-driven robotic systems.
- Familiarity with modern RL environments and physics simulators, or sim-to-real techniques such as Domain Randomization, System ID-based digital twins, or Actuator Net modeling.
Tools
Reinforcement Learning (RL), Quasi-Direct Drive (QDD), MuJoCo, Isaac Sim/Gym, Drake, Bullet, Domain Randomization, Actuator Net
AI focus
Applies reinforcement learning to robotic control and develops simulation-to-hardware transfer for RL policies.
Model-written summary · the original posting is authoritative