Job Description

Meta is seeking Research Scientist Interns to join a research team in Singapore working on real-time audio-visual interaction and generation. The team builds models that understand and generate media from text, audio, and images, and serves them interactively at low latency. The role sits at the intersection of multimodal understanding, agentic generation, and efficient modeling. Interns work on open research problems with a direct path to product, and are encouraged to publish their results.

Responsibilities

Research and develop audio-visual interactive and generative models, such as any-to-any models across speech, text, and video, real-time agentic interactive systems, and autoregressive generation.

• Improve inference speed and quality through distillation, streaming generation, and efficient architectures.
• Develop post-training methods, including fine-tuning and multimodal reinforcement learning.
• Build evaluation methods and benchmarks for long-form, multi-turn interaction and generation.
• Work with researchers and engineers to bring results into prototypes, and publish findings at peer-reviewed venues.

Qualifications

Currently pursuing a PhD in Computer Science, Electrical Engineering, or a related field

• Research experience in at least one of: generative models, speech-to-speech (S2S) models, or reinforcement learning
• Experience with Python and a deep learning framework such as PyTorch First-author publications at venues such as CVPR, ICCV, ECCV, NeurIPS, ICLR, or ICML
• Experience with reinforcement learning for generative models, or with LLM-based agents
• Experience training large models on distributed GPU clusters
• Experience building or post-training frontier-scale, high-quality models
• Strong interest in next-generation interactive models and agentic workflows