Research Scientist Intern: Embodied AI & Robotics

Meta; Menlo Park (CA); On-site; USD 55,000 - 83,000; Part time

Job summary

Meta Reality Labs Research is seeking a Research Scientist Intern to develop scalable mid-training methods for world action models (WAMs) and vision-language-action models (VLAs) for robotic manipulation. You will collaborate with researchers and engineers to advance embodied AI on large-scale egocentric video and robot data.

Internships span 12 to 24 weeks with multiple start dates; you will design experiments, build data pipelines, and publish results at top venues in ML, vision, and robotics.

Qualifications

• Pursuing or has a PhD in CS/ML/AI/Robotics.
• Experience training video models or robotics foundation models.
• Experience with Python and PyTorch.
• Experience designing controlled experiments and debugging ML algorithms.
• Experience with VLAs, WAMs or embodied decision-making is preferred.
• Experience with distributed training across GPUs/nodes.
• Proven research track record (papers/grants/patents).

Responsibilities

• Develop scalable mid-training methods for WAMs using large-scale data.
• Investigate representation and objectives for learning from human egocentric video.
• Address challenges in training on web-scale data, including embodiment and domain gaps.
• Build efficient training and data pipelines and study scaling effects.
• Design experiments and benchmarks to measure mid-training contributions.
• Collaborate with researchers and engineers to evaluate models in simulation and on real robots.
• Publish research findings at leading ML, CV, and robotics venues.

Skills

Python

PyTorch

Experiment design

Distributed training

Robotics experience

Research publications

Education

Ph.D. in Computer Science/ML/AI/Robotics

Job description

Meta Reality Labs Research is seeking a Research Scientist Intern to develop scalable mid-training methods for world action models (WAMs) and vision-language-action models (VLAs) for robotic manipulation. You will collaborate with researchers and engineers to advance embodied AI on large-scale egocentric video and robot data.

Internships span 12 to 24 weeks with multiple start dates; you will design experiments, build data pipelines, and publish results at top venues in ML, vision, and robotics.