Research Scientist Intern - On-Device AI Model Optimization

Meta; Menlo Park (CA); On-site; USD 28,000 - 50,000; Full time

Job summary

Reality Labs is seeking interns to develop efficient machine learning models for AR/VR applications. You will work on computer vision, speech, sensor fusion, NLP, and multimodal AI within tight compute and power constraints. The role offers hands-on exposure across the stack, from algorithms to hardware deployment.

Internships range from 12 to 24 weeks with start dates throughout the year. You will collaborate with experts and partner with software and hardware teams to bring state-of-the-art

Qualifications

• Background in developing ML models for AR/VR applications.
• Experience delivering real-time on-device models.
• Familiarity with on-device optimization techniques.

Responsibilities

• Work across the full stack from user experiences and algorithms to hardware execution.
• Collaborate with domain experts to understand application challenges and develop neural networks.
• Assist in deploying on-device solutions with software and hardware teams.

Skills

Machine learning

Computer vision

Speech

Sensor fusion

NLP

Multimodal AI

On-device deployment

Real-time models

Job description

Reality Labs advances Meta's vision for virtual reality (VR) and augmented reality (AR). Delivering compelling experiences on AR/VR devices requires innovation and co-design across the full technology stack, from algorithms to custom silicon. Our teams push the state of the art in computer vision, speech, virtual assistants, machine learning, mixed reality, graphics, displays, sensors, and human body tracking.

We are seeking interns with a background in developing efficient machine learning models for AR/VR applications, including computer vision, speech, sensor fusion, natural language processing (NLP), and multimodal AI. Our team develops high-performing models within tight compute, memory, and power constraints, using optimization techniques such as sparsity, quantization, compression, neural architecture search, and tensor decomposition. Ideal candidates will have practical experience building real-time models that maintain high accuracy under on-device deployment constraints.

In this role, you will gain experience across the full stack, from user experiences and algorithms to hardware execution. You will collaborate with domain experts to understand application challenges and develop state-of-the art neural networks to address them, then partner with software and hardware teams to deploy these solutions on-device.

Our internships range from twelve (12) to twenty-four (24) weeks, with start dates throughout the year.