Robotics Simulation and Navigation Engineer — Real-to-Sim

mid
Other
2 views0 saves0 applied

Quick Summary

Key Responsibilities

mesh extraction, collision and material setup, semantic labels, USD export, with the venue’s furniture and lighting represented well enough to matter.

Requirements Summary

RWS Holdings PLC does not accept agency resumes. Please do not forward any unsolicited resumes to any RWS employees. Any unsolicited resume received will be tre

Technical Tools
Other

About the Role

~2 min read

 

You own the loop between the real venues our robots work in and their simulated twins.

Our humanoids interact with people in real venues: hotels, clinics, museums, offices. Today the robot stands still and gives directions from a tagged map. The next step is for it to move with people, and before a humanoid walks among strangers we want to have tested that behavior in a faithful simulation of the exact venue. The step after that is to use those twins to evaluate robot policies and behaviors at a scale no number of live sessions can reach, and to offer venue twins and sim-based evaluation to the robot makers we work with alongside the real data we collect there.

You build that capability: capture real environments, turn them into simulation-ready twins using the real-to-sim tooling that now exists rather than writing your own, make the robot navigate in the twin and then in the real venue, and close the gap between the two. This is a long-term role that starts with one venue and one robot and grows into a capture-to-twin-to-evaluation pipeline used across our physical AI work.

What this role is not: you will not develop a physics engine or a simulator core, and you will not build general simulation infrastructure. We adopt Isaac Sim-class platforms and the open real-to-sim pipelines around them; your job is to make them work for our venues and our robots

 

 

RWS TrainAI is the AI data and evaluation business of RWS Group, working with leading AI developers worldwide. We are building a physical AI practice: multilingual, culturally localized interaction for service robots, human evaluation of how robots behave around people, and the data pipelines behind both. Our advantage is a global, multilingual human network and a focus on behavior that is culturally appropriate, measured and auditable.

 

Responsibilities

~1 min read
  • Stand up and adapt existing real-to-sim tooling (Isaac Sim with NuRec, Genie Sim, Gaussian-splat and photogrammetry pipelines, handheld LiDAR scanners) into a venue capture pipeline, aligned to the tagged point-of-interest map our deployments already use.
  • Turn reconstructions into simulation-ready scenes: mesh extraction, collision and material setup, semantic labels, USD export, with the venue’s furniture and lighting represented well enough to matter.
  • Populate twins with pedestrian agents (crowd simulation) matched to the venue’s traffic, so social navigation can be tested against people, not empty corridors.
  • Define what “faithful enough” means for each use (navigation testing, perception testing, policy evaluation) and measure it.
  • Build localization and navigation for our humanoid in human environments: LiDAR-inertial SLAM (Livox-class sensors, FAST-LIO2 / Point-LIO or equivalent), Nav2 or equivalent planning, dynamic obstacle avoidance among pedestrians, and the social rules (keep distance, give way, slow near groups) our HRI scientist specifies.
  • Test every navigation change in the venue twin before it runs in the venue, with the systems integration engineer who owns the robot and its safety envelope.
  • Integrate with the application layer so a walk through the venue is a sequence of points of interest the dialogue can narrate, and build the recovery behaviors for when navigation fails in front of people. The robot runs in production; you share the on-call rotation for it.
  • Emulate our robot’s sensors in the twin (RGB, depth, LiDAR, microphones where it matters) closely enough that perception results transfer, and calibrate dynamics between twin and robot.
  • Close the gap: domain randomization, the measurements that tell us when a result in simulation predicts a result in the venue, and a tracked sim-to-real gap for navigation.
  • Build closed-loop and log-replay evaluation of robot behaviors and policies in the twin, replaying our real interaction and demonstration data, so that behavior changes and third-party policies can be tested at scale before live sessions.
  • Work with our data and research engineers so the twins, the real recordings and the evaluation results share one schema and can be delivered together to robot makers.

 

Requirements

~1 min read

 

 

Get the 3Ps right – Partner, Pioneer, Progress – and we´ll Deliver together as RWS.

 

Recruitment Agencies: RWS Holdings PLC does not accept agency resumes.  Please do not forward any unsolicited resumes to any RWS employees.  Any unsolicited resume received will be treated as the property of RWS and Terms & Conditions associated with the use of such resume will be considered null and void.

  • Degree in robotics, computer science or a related field, or equivalent experience.
  • 3+ years of robot simulation: operating and extending an Isaac Sim / Isaac Lab, MuJoCo or Genesis-class platform for training or evaluation, including building scenes and assets (USD or equivalent) and a real-to-sim pipeline of some kind.
  • 3+ years of navigation on real mobile robots in human environments: indoor localization and mapping, 3D sensors for world modeling, planning and avoidance around people.
  • Strong Python and C++; ROS 2; LiDAR and IMU fusion; comfortable with GPU workflows.
  • Experience taking something from simulation onto a real robot and dealing with what broke, and with closed-loop or log-replay evaluation.
  • Professional working English; Mandarin a strong plus for working with venues and the China team.

Nice to Have

~1 min read
  • Hands-on neural scene reconstruction (3D Gaussian splatting, NeRF) or photogrammetry, and turning it into simulator assets.
  • Navigation or locomotion on a legged or humanoid platform.
  • Sensor emulation, domain randomization, and numerical optimization for calibration or motion planning.
  • Social navigation research or deployment in retail, hospitality, healthcare or museums.
  • Semantic or open-vocabulary mapping (scene graphs) for linking the navigation map to a knowledge base.
  • Experience with closed-loop or log-replay evaluation of robot policies at scale, or with VLA / world-model training data.

 

  • First 3 months: a live demo in front of an audience. Not polished: the first venue captured and reconstructed into a navigable simulation twin, the humanoid localizing and navigating in it with pedestrian agents, shown side by side with the real venue; capture protocol documented so a second venue takes a week.
  • By 6 months: the humanoid navigating in the real venue with the social rules in place, every change tested in the twin first; a measured sim-to-real gap for navigation with a plan to close it.
  • By 12 months: three venues captured and simulated; sim-based evaluation running on our real data; a first venue twin plus evaluation report delivered to a robot maker.

 

Location & Eligibility

Where is the job
Not specified
Location terms not specified

About this job

Posted
1 day ago
Employer
has 27 open jobs on Jobera.

Listing Details

Posted
November 18, 2026
First seen
November 19, 2026
Last seen
November 19, 2026

Posting Health

Days active
-1
Repost count
0
Trust Level
49%
Scored at
November 19, 2026

Signal breakdown

freshnesssource trustcontent trustemployer trust
Newsletter

Stay ahead of the market

Get the latest job openings, salary trends, and hiring insights delivered to your inbox every week.

A
B
C
D
Join 12,000+ marketers

No spam. Unsubscribe at any time.

Robotics Simulation and Navigation Engineer — Real-to-Sim