Humanoid Robotics Engineer: Whole-Body Control & Estimation

S27a; Paris; Hybride; EUR 70 000 - 100 000; Plein temps

Résumé du poste

UMA builds general-purpose humanoid robots. Founded in Paris in 2025 by alumni of Tesla Optimus, Google DeepMind, and LeRobot, we're building the humanoid and its control stack from the ground up.

We're hiring a robotics engineer to own the whole-body controller and state estimation on our humanoid. You will design and tune the controller for locomotion, build real-time state estimation, and solve MPC/Inverse dynamics problems on real hardware.

Qualifications

• Expertise in whole-body control on real hardware (QP inverse dynamics, MPC, task-space or hierarchical control)
• Strong state estimation background (floating-base, contact detection, sensor fusion)
• Proficiency in C++ and real-time systems (PREEMPT_RT, RTOS)
• Rigid-body dynamics and numerical optimization (Pinocchio, Crocoddyl, or equivalent)
• Experience bringing up a legged or humanoid robot
• Bonus: fluency in MuJoCo or Isaac Sim, loco-manipulation and force/impedance control

Responsabilités

• Design and tune the whole-body controller for locomotion and loco-manipulation
• Build state estimation on the real robot: floating-base estimation, contact detection, sensor fusion
• Formulate and solve real-time optimal control problems (QP inverse dynamics, MPC)
• Characterize actuator behavior and design controllers within real hardware limits
• Combine model-based control with learned policies at the model/learning seam

Connaissances

whole-body control

state estimation

C++

real-time systems

rigid-body dynamics

numeric optimization

Outils

Pinocchio

Crocoddyl

MuJoCo

Isaac Sim

Description du poste

UMA builds general-purpose humanoid robots. Founded in Paris in 2025 by alumni of Tesla Optimus, Google DeepMind, and LeRobot, we're building the humanoid and its control stack from the ground up.

We're hiring a robotics engineer to own the whole-body controller and state estimation on our humanoid. You will design and tune the controller for locomotion, build real-time state estimation, and solve MPC/Inverse dynamics problems on real hardware.