AI Research Scientist - Multimodal Vision & On‑Device Models

Meta; Redmond; On-site; AUD 174,000 - 257,000; Full time

Benefits offered by this job

Bonus

Equity

Benefits

Job summary

Meta Reality Labs is seeking an AI Research Scientist to advance vision and multimodal perception. You will build models that understand images and videos, work with efficient encoders and multimodal LLMs, and define capabilities for the next generation of AI.

The role involves training and benchmarking state-of-the-art methods while collaborating with diverse teams. You will also handle large datasets, contribute to data annotation efforts, and publish impactful results.

Qualifications

• Bachelor's degree or equivalent practical experience in CS/related field.
• PhD in AI/vision/robotics or related area preferred.
• Experience with deep learning, computer vision, robotics or AI infrastructure.
• Proficiency in Python and PyTorch.
• Experience developing software and running complex experiments.
• Ability to communicate research to peers.

Responsibilities

• Build models that understand images and videos.
• Research with efficient vision encoders and multimodal LLMs.
• Define, build and benchmark capabilities for next-gen AI.
• Train and optimize state-of-the-art ML and neural networks.
• Work with large datasets and contribute to data-annotation engines.

Skills

Python

PyTorch

Deep Learning

Computer Vision

Research Communication

Education

Bachelor's degree in Computer Science or related field

PhD in AI/related field

Tools

Git

Jupyter

Job description

Meta Reality Labs is seeking an AI Research Scientist to advance vision and multimodal perception. You will build models that understand images and videos, work with efficient encoders and multimodal LLMs, and define capabilities for the next generation of AI.

The role involves training and benchmarking state-of-the-art methods while collaborating with diverse teams. You will also handle large datasets, contribute to data annotation efforts, and publish impactful results.