Hiring nowSimulation, digital twins and physical AI assurance

Sim-to-Real Researcher

Also called Simulation Transfer Scientist · Sim-to-Real Engineer · Robotics Simulation Engineer

A Sim-to-Real Researcher cuts the failures that appear when a policy trained in simulation meets a physical system. It matters because a robot that only works in simulation is a demo, not a product.

What they do

Takes policies that perform well in simulation and works out why they fail on hardware. Builds and tunes the simulated environment, then applies domain randomisation, system identification and adaptation methods to close the gap. Designs real-world test runs that compare methods fairly, with repeatable setups and logged disturbances. Works closely with controls, perception and hardware engineers to trace faults to their source, whether that is contact modelling, latency, sensor noise or actuator behaviour. Writes up what worked, what did not, and what the team should simulate next.

What a strong hire delivers

Within the first months, a strong hire gives the team a transfer method that is documented and reusable, not a one-off fix. They set up a real-world comparative evaluation that shows which approach holds up. They produce a clear account of the main sources of the gap, and a prioritised plan for narrowing it. Policies fail less often on hardware, including under disturbances nobody planned for.

How you can tell they are good

Ask them to walk through a policy that worked in simulation and failed on hardware. A strong candidate names the cause, shows how they confirmed it, and explains what they changed. Ask how they chose the randomisation ranges and how they tested against unseen disturbances. Look for real hardware results, honest comparisons against a baseline, and plain explanations of what remains unsolved.

Why it is hard to fill

Very few people have done this end to end. The work sits between simulation, control and robotics hardware, and most candidates have only two of the three. Titles vary widely, so searching by title misses good people. Many come from graphics or physics simulation and have never faced a real robot. Others know hardware well but have limited experience of learned policies. Employers compete for the same small group.

Where to find them

Look at robotics companies and research groups working on learned control and manipulation. Strong candidates often hold titles in simulation engineering, controls, robotics research or systems verification. Others come from physics simulation, graphics or game engines and have moved toward physical systems. Digital-twin teams in industrial and autonomous systems firms are another source, especially people who have validated models against measured behaviour.

Indicative pay

Not enough published evidence yet to give a fair range for this role, which is common for work this new. We benchmark it for you as part of a search.

Interview questions for a Sim-to-Real Researcher

  1. Tell me about a policy that worked in simulation and failed on hardware. What was the cause?

    A strong answer names a specific cause, such as latency, contact modelling or sensor noise, and explains how it was confirmed with evidence. Worry if the answer is vague, blames hardware generally, or has no real-world result.

  2. How do you decide which parameters to randomise and how wide the ranges should be?

    Good answers link ranges to measured data or system identification and discuss the cost of over-randomising, which can make policies timid. Worry if they randomise everything widely and hope, without a way to check.

4 more questions, with what a strong answer covers, are in the Role Kit.

Job description you can post

We build robotic systems that learn in simulation and must work reliably in the physical world. We are hiring a Sim-to-Real Researcher to reduce the failures that appear when policies move onto hardware, and to prove that the improvement holds up in real tests.

What you will do

  • Diagnose why policies that succeed in simulation fail on physical systems
  • Design and improve simulated environments so they better match real behaviour
  • Develop and apply domain randomisation and adaptation methods
  • Plan and run real-world comparative evaluations of transfer approaches
  • Test performance under disturbances not seen in training
  • Work with controls, perception and hardware engineers to trace faults
  • Document methods and findings so others can reuse them

What you will bring

  • Demonstrated work moving a learned or controlled policy from simulation to hardware
  • Hands-on experience with physics simulators and robotics control
  • Evidence of designing fair, repeatable real-world experiments
  • Ability to trace a failure to its cause using logged data
  • Clear written and spoken communication of results and limits
  • Comfort working with hardware as well as code

Nice to have

  • Experience with system identification
  • Background in graphics, physics simulation or verification
  • Experience with digital-twin platforms

Free to use and adapt. If it saves you time, a link back to this page is appreciated.

Skills that matter

  • Simulation and digital twinsAdvancedBuild useful virtual representations of environments
  • Robotics and controlAdvancedPerceive, plan and act through physical systems
  • Model evaluationWorkingMeasure capability and failure modes credibly
  • Technical communicationWorkingExplain evidence and limitations to relevant readers

Other roles in simulation, digital twins and physical AI assurance

Hiring nowRobotics Simulation EngineerA Robotics Simulation Engineer builds virtual environments where robots can be developed and tested, then proves those results hold on real…Hiring nowDigital Twin EngineerA Digital Twin Engineer keeps a virtual model of a physical asset or process accurate and useful. The company gets predictions it can trust for…EmergingPhysical AI Validation EngineerA Physical AI Validation Engineer tests whether learned robot behaviour stays safe at its limits. They give the company evidence to decide if a…

Related roles in other families

Hiring nowRobot Learning EngineerRobot learning and manipulationA Robot Learning Engineer trains policies that let robots complete defined physical tasks reliably. The work turns a promising demo into behaviour a…Hiring nowRobot Manipulation ResearcherRobot learning and manipulationA Robot Manipulation Researcher builds the grasping and object-interaction methods that let robots handle real things reliably. They turn lab demos…EmergingVision-Language-Action ResearcherRobot learning and manipulationA Vision-Language-Action Researcher builds models that turn instructions and camera input into safe, bounded robot actions. They matter because they…Hiring nowAutonomy Perception EngineerAutonomous navigation and field roboticsAn Autonomy Perception Engineer turns raw sensor data into a reliable picture of the world that a machine can act on. It decides whether an…

How we wrote this

Role definitions are our own editorial work, checked against occupational classifications, standards, research and employer pages. Titles vary by company, and this is a description of the work, not a job advert.

← Simulation, digital twins and physical AI assuranceAll new job families

Start a search

Hiring a Sim-to-Real Researcher?

Tell us the work. We find the people already doing it, whatever their job title says, and you meet them already interviewed for your role.