Persons in these roles are expected to work from our offices in Seattle. On-site requirements vary based on position and team. If you have questions about on-site work arrangements for this role, please ask your recruiter.

Our base salary range is $128,880 - $193,320, and in addition we have generous bonus plans to provide a competitive compensation package.

Who You Are

You will own the control stack on our physical robots and the path from simulation to them. This is first and foremost a controls role: you will bring up and tune real-time, torque-level control on arms, multi-fingered hands, bimanual systems and humanoids, make it stable and safe enough that researchers run experiments on hardware every day, and build the interfaces that learned policies plug into. Experience with humanoid whole-body control and with Unitree platforms (G1, H1 and similar) is a strong plus, as humanoid and whole-body work is a growing share of the lab. You will also design and build the hardware the research needs, from end-effectors to test rigs. You will work closely with research scientists and the simulation team so that what is learned in simulation shows up as reliable behaviour on hardware.

Who We Are

The Robotics team at Ai2, led by Dieter Fox, uses scalable simulation to study how foundation models scale and generalize for robotics. We focus on manipulation, humanoid and whole-body control, large-scale simulated training, sim-to-real transfer, robot foundation models and skill learning. We publish at CoRL, ICRA, RSS, CVPR, NeurIPS, ICLR and ICML, and everything we build is open source.

Your Next Challenge

Build and own the real-time control stack: joint, impedance, operational-space and torque control on torque-controlled arms, hands and humanoids, with motion planning, safety layers and the interfaces that learned policies plug into, across embodiments.
Bring up, tune and stabilize whole-body controllers for humanoids and bimanual systems: QP-based or MPC-based whole-body control, state estimation (IMU, encoders, force/torque), contact handling and fall or fault recovery, and the sim-to-real transfer of learned locomotion and loco-manipulation policies.
Lead the setup, calibration, maintenance and tuning of our robots: arms, grippers, multi-fingered hands, humanoids, cameras and tactile sensors, across single-arm, bimanual, mobile and legged platforms.
Diagnose stability, latency and contact problems across hardware, firmware, control and learned policy, using system identification and dynamics models to turn algorithms into reliable real-world behaviour.
Design and build the hardware the lab needs: end-effectors, fingertips, sensor mounts, fixtures and test rigs, from CAD through 3D printing or machining, including integrating cameras, tactile sensors and hands.
Build the runtime and tooling around the robots: teleoperation and data collection, logging, monitoring, deployment and operator-facing tools, so the platform scales across robots, tasks and people.
Contribute to open-source releases and publications.

What You’ll Need

Minimum Qualifications:

Master’s with 3+ years of experience, or PhD, in robotics, control or a related field. Strong candidates with a bachelor’s and 3+ years of industry experience are also considered.
Controls: you have implemented and tuned real-time controllers (joint, impedance, operational-space or torque control) on physical robots, with a solid grounding in rigid-body dynamics, control theory, stability analysis and system identification. You can explain why a controller went unstable and fix it.
Whole-body and floating-base control: you understand floating-base dynamics, contact constraints and inverse dynamics, and have worked with at least one of QP-based whole-body control, operational-space control or MPC, on hardware or in a high-fidelity simulator.
Manipulation: substantial hands-on experience with real-robot manipulation, including bring-up, calibration and control of arms with grippers or hands.
System design: you have designed robot software systems end to end (control loops at 500 Hz to 1 kHz, state estimation, data flow, and the interfaces between planning, perception and learning) and can reason about latency, reliability and failure modes. Strong Python; solid real-time C++.