Research Intern, Robotics Data & Real-to-Sim

WorldEngine AI; San Francisco (CA); On-site; USD 34,000 - 55,000; Full time

Job summary

WorldEngine AI in San Francisco seeks a Research Intern to turn real-world observations into high-quality datasets and simulation-ready environments.

You will collaborate with engineering and research teams on 3D reconstruction, real-to-sim workflows, and experiments that measure data quality impacts on robot learning and evaluation.

Qualifications

• Pursuing a degree in CS, robotics, EE, or equivalent hands-on research.
• Strong Python and ML/CV experience.
• Background in 3D geometry or robot learning.
• Ability to work with imperfect real-world data and design experiments.

Responsibilities

• Develop real-to-sim workflows turning real images, video, and sensor data into simulation-ready scenes.
• Build tools to process, align, and validate robotics data including 3D geometry and trajectories.
• Investigate how reconstruction fidelity, scene coverage, and physical properties affect learning.
• Design experiments to compare reconstructed environments with real data and measure impact on learning.
• Turn research prototypes into reproducible data pipelines and evaluation tools.
• Document methods and findings for reports or research publications.

Skills

Python

ML

CV

Research

Education

CS/Robotics EE degree

Job description

About the Role

At WorldEngine, we’re building the data infrastructure for Physical AGI. Real-world experience is central to that effort: the scenes robots operate in, the objects they interact with, and the data that captures what happens.

As a Research Intern, you’ll investigate how real-world observations and robot interactions can be converted into high-quality datasets and simulation-ready environments. You’ll work closely with our engineering and research teams on 3D reconstruction, real-to-sim workflows, and experiments that measure how data and simulation quality affect robot learning and evaluation.

What You’ll Do

• Develop real-to-sim workflows that turn real-world images, video, and sensor data into simulation-ready scenes, objects, and environments

• Build tools to process, align, and validate robotics data, including visual observations, 3D geometry, and robot trajectories

• Investigate how reconstruction fidelity, scene coverage, and physical properties affect the usefulness of reconstructed environments for robot learning

• Design experiments and benchmarks to compare reconstructed environments with real-world reference data and measure their impact on robot learning and task performance

• Work with engineers to turn research prototypes into reproducible data pipelines, datasets, and evaluation tools

• Document methods, experiments, and findings, and contribute to technical reports or research publications where appropriate

What We’re Looking For

• Currently pursuing a degree in computer science, robotics, electrical engineering, or a related field, or equivalent hands‑on research experience

• Strong Python programming skills and experience implementing and testing machine learning or computer vision methods

• A foundation in 3D geometry, computer vision, or robot learning, with the ability to turn research ideas into working prototypes

• Comfortable working with imperfect real-world data, debugging pipelines, and designing clear, controlled experiments

• Clear technical communication and the ability to take ownership of a scoped research project in an early-stage team

Nice to Have

• Research or project experience in 3D reconstruction, neural rendering, or scene understanding

• Experience with robotics simulators or tools for scene and asset generation

• Hands‑on experience training or evaluating robot policies in simulation or on physical systems

• Experience building reproducible experiments, processing multimodal datasets, or contributing to open‑source projects