Job Description

Meta Reality Labs Research (RL Research) is a team of researchers and engineers pushing the frontier of AI, robotics, and AR/VR technology. Within RL Research, our team is building the next generation of foundation models (VLAs, WAMs) for robotic manipulation. We are looking for a skilled Research Engineer to take those models onto real hardware in our lab.

This role covers the full robotics stack: hardware bring-up, teleoperation, data collection, and running learned policies on real robots. It involves building prototypes together with research scientists and engineers, running the experiments that show whether they work, and finding the problems that only appear once a model has to act in the physical world.

Responsibilities

Build, operate, and maintain robot platforms for teleoperation and deployment

• Deploy learned policies on real robots and optimize their performance on-robot
• Work with researchers to design experiments and measure how well policies work on real hardware
• Debug failures across hardware, perception, data, and policies
• Turn research results into working prototypes
• Build tooling that makes experiments repeatable
• Publish research results in top conferences and release code that contributes to the field of robotics

Qualifications

Currently has, or is in the process of obtaining, a PhD degree in robotics, computer vision, machine learning, or a related field

• Industry experience with modern, general-purpose robot platforms
• Hands-on experience bringing up, calibrating, and debugging a new robot platform from scratch
• Experience running learned policies on real robots and improving their performance
• First-author publications at ICRA, CoRL, RSS, IROS, or CVPR, or widely used open-source contributions
• Proficient in Python and PyTorch Contributed to open-source robot learning projects such as LeRobot, robosuite, or ManiSkill
• Trained or evaluated large 3D generative or world models
• Built teleoperation rigs and used them to collect robot data at scale
• Debugged policy failures across hardware, data, training, and simulation
• Worked on long-horizon planning for robots
• Worked on 3D computer vision for robots, such as reconstruction, pose estimation, or scene understanding
• Worked on contact-rich manipulation, including force or tactile feedback
• Trained or fine-tuned vision-language-action models or world action models
• Fluent in ROS or ROS 2, including kinematics and camera-to-robot calibration