←

Robotics Deployment Engineer

Shenzhen · Customer travel requiredFull-timeOn-siteRMB 25,000–40,000 / month中文

About the Role

Intuition is a Silicon Valley robotics startup backed by some of the most prestigious investors in tech, connected to Founders Fund, MiniMax, Raspberry Pi, and others. Our team has backgrounds from Stanford University, UC Berkeley BAIR, Fudan AI, CMU Robotics Institute, and Google, and we work with industrial partners including KUKA, UR, Midea, Mercedes Benz, P&G, IKEA, and SF Express Logistics.

You'll join the Rapid Prototyping team at our Shenzhen deployment center: robots from many vendors arrive with an application, we build the station platform around them, and the solution ships. Every station shares one platform: a height-adjustable torso, two arms at ALOHA spacing, a top-view neck camera, fish-eye wrist cameras, a mid-height torso camera, and one Jetson Orin. V0 is a stationary single-arm station; V1 adds the second arm, one cable to the wall, and wheels that lock solid; V2 goes mobile with wheeled navigation, battery, and 5G.

You'll integrate everything into the robot, and you're the reason the model runs perfectly. You're the jack of all trades on the station. Manage the teleoperators. Collect the data. Evaluate the model. Find out why the robot failed. Fix the pipeline. Train the next version. Put it back on the robot.

The stack you inherit is real. A GELLO leader arm runs at 50 Hz while the industrial arm's real-time control loop expects commands at 300 Hz. Open-source Ruckig interpolation closes that gap, hand-tuned to the arm's velocity, acceleration, and jerk limits. A cold-start handshake waits a few cycles before the first command is written. The stack already runs KUKA over RSI, plus AgileX Piper, ARX X5, and YAM over CAN-to-USB. You'll take on the next vendor's SDK, including flange IO and the separately wired gripper, and fold that vendor's arm into a single deployment platform.

At our stage, deployment is research. What happens on the factory floor determines what data we collect, what models we train, and what infrastructure we build next. You'll work directly with the CTO and founding team, and the scope is intentionally broad. Whoever sees a failure in the field should be able to trace it across the complete system, from teleoperator to environment, and fix what broke. This is an early, high-ownership role, not a narrowly defined graduate position.

What You'll Do

  • Integrate everything into the robot: arms, gripper, four USB cameras, Jetson Orin, and the model, until the station runs its application.
  • Own the leader/follower teleoperation stack: bridge a 50 Hz GELLO leader arm to a 300 Hz industrial arm without tripping a failsafe.
  • Get through vendor SDKs (KUKA RSI, flange IO, CAN-to-USB arms) and unify them into a single deployment platform.
  • Solve the inverse kinematics for every arm that lands on the station.
  • Make the model run ultra-smooth on its dedicated task with real-time action chunking and interpolation.
  • Schedule model APIs, orchestrate the model running on the robot, and implement local recovery modes like DAgger.
  • Run and validate data collection at partner factories: teleoperator management, task design, and data quality.
  • Debug a robot fast: diagnose failures across teleoperator, task design, dataset, model, software, hardware, and environment.
  • Evaluate VLA models on real robots, fine-tune with the research team, and redeploy to verify the gains on hardware.

What We're Looking For

  • No hard prerequisites to apply: we hire on what you have built, not on titles or experience with our exact stack.
  • A jack of all trades: comfortable moving between software, models, robots, and operational problems.
  • Grounded in the principles of robot learning and in how a leader/follower teleoperation stack works.
  • Strong software and computer engineering fundamentals, with evidence you can independently understand difficult systems.
  • You understand how complex real deployment environments are, and when nobody has the answer you teach yourself one.
  • Hands-on robotics experience: RoboMaster, Robocon, a robotics team, a research lab, or systems you built yourself.
  • Based in Shenzhen, willing to travel regularly to partner factories, and fluent in English.
  • Bonus: you have driven an industrial arm through its vendor SDK, or a light arm like AgileX Piper over CAN-to-USB.
  • Bonus: you have gone below a vendor SDK on a real arm and tuned its control loop, PID gains, or velocity profile.
  • Bonus: agentic coding tools like Claude Code or Codex are already part of your workflow.

Why Join Us

You'll own the complete loop from teleoperation to model training in your first months. Direct mentorship comes from UC Berkeley BAIR professors and researchers at Physical Intelligence. Every station that ships from the deployment center runs because you made it run. Compensation is RMB 25,000–40,000 per month, plus annual bonus and meaningful early-employee equity. We can assist eligible candidates with work-permit arrangements.