Research Scientist, Robotics – World Models

InnoData

. Define how Innodata designs, structures, and evaluates data for robot foundation models .

Posted 9/24/2026full-timeRemote • United StatesMid-LevelSenior💰 $160,000 - $185,000 per yearWebsite

Core Competencies

Role fit

Core Competencies

Use this summary to align your resume positioning with the role.

Demonstrates expertise in defining and evaluating data for robotics foundation models, with a strong focus on data specifications, evaluation methodologies, and real-world transfer. Proficient in curating and managing heterogeneous robot datasets while collaborating effectively with cross-functional teams.

Highest-signal resume keywords

Robot Learning ExperienceStrong PyTorch FundamentalsNVIDIA Isaac Sim ExperienceImitation Learning and Reinforcement LearningFirst-Author Publications

ATS Keywords

Tailor your resume

Applicant Tracking System Keywords

Tip: use these terms in your resume and cover letter to boost ATS matches.

Hard Skills

Data SpecificationData CurationEvaluation MethodologiesImitation LearningReinforcement LearningDomain RandomizationSystem IdentificationSim-to-Real TransferData Quality AblationsScaling Studies

Soft Skills

Clear CommunicationCollaborative TeamworkRigorous Documentation

Tools & Technologies

NVIDIA Isaac SimIsaac LabOmniverseMuJoCoHuggingFace TransformersPEFT

Industry Keywords

Robot Foundation ModelsMotion-Capture DataEgocentric Data CollectionExocentric Data CollectionSynthetic Data PipelinesLeRobot Dataset FormatRLDSOpen X-EmbodimentAction RepresentationsBenchmarking Methods

Tech Stack

Tools & technologies

PyTorch

About the role

Key responsibilities & impact

Define how Innodata designs, structures, and evaluates data for robot foundation models
Translate robotics foundation-model requirements into data specifications, including modalities, action representations, sampling, annotation schemas, and evaluation criteria
Decide what to capture in the real world versus generate in simulation
Curate and weight training mixes across heterogeneous robot datasets, embodiments, action spaces, and sensor setups
Guide data collection across motion-capture, egocentric, exocentric, teleoperation, multi-sensor, and synthetic-data pipelines
Build evaluation and benchmarking methods that predict real-world transfer, including world-model evaluation and sim-to-real methodology
Fine-tune and evaluate foundation models on Innodata data through data-quality ablations and scaling studies
Design adversarial and long-horizon evaluations and convert failure modes into improved data
Publish benchmarks, methodologies, and papers
Collaborate with capture, annotation, and synthetic-data teams to operationalize collection and labeling plans

Requirements

What you’ll need

Roughly 4+ years of hands-on industry experience in robot learning or robotics ML
Bachelor's degree in computer science, electrical engineering, robotics, or a related technical field is required
Trained and evaluated robot policies using imitation learning or reinforcement learning
Strong PyTorch fundamentals
Experience curating, filtering, and weighting robot data across embodiments and sensors
Fluency in the LeRobot dataset format, RLDS, Open X-Embodiment, and common motion and sensor formats
Hands-on experience with NVIDIA Isaac Sim, Isaac Lab, Omniverse, MuJoCo, or comparable simulation engines
Experience with domain randomization, system identification, and sim-to-real transfer
Experience with teleoperation or egocentric data collection
Experience adapting VLM backbones for control and fine-tuning large VLA models using HuggingFace transformers and PEFT
First-author publications or strong open-source contributions at venues such as CoRL, ICRA, IROS, RSS, or NeurIPS
Ability to work with customer and frontier-lab research scientists and explain data and modeling decisions clearly
Rigorous, reproducible approach to experiments and documentation