Research Scientist Intern, FAIR - Video Generation Self-Improvement (PhD)

Meta•Seattle, WA

9d•$7,650 - $12,667

About The Position

Meta is seeking a PhD Research Intern to join the Fundamental AI Research (FAIR) team within Meta Superintelligence Labs. The role focuses on advancing video generation, post-training of large-scale models, and self-improving agentic systems, helping build perceptual and generative foundations for next-generation AI systems. Interns collaborate with scientists, engineers, and cross-functional partners and may make core algorithmic advances using Meta’s research resources and facilities. Internships last 12 to 24 weeks, with various start dates throughout the year. The research areas of interest include generative modeling for video, such as diffusion, flow matching, and video synthesis; post-training of large-scale video and multimodal models, including supervised fine-tuning, reinforcement learning, optimization, and data curation; multimodal understanding, reasoning, and generation; self-improving and agentic systems; and reward modeling and LLM/VLM-as-a-judge. Interns will conduct fundamental and applied research, develop and test novel methods, analyze results, document and present research outcomes, and contribute research that can be applied to Meta products. The role also involves collaboration across Meta on perceptual foundations for real-time embodied agents and conversational AI, and contributions to frontier models such as Llama and large multimodal models.

Requirements

Currently has or is in the process of obtaining a PhD in Computer Science, Electrical Engineering, Computer Vision, Machine Learning, Artificial Intelligence, a relevant technical field, or equivalent.
Research and/or work experience in video generation, computer vision, or multimodal AI.
Experience with large-scale model training, post-training optimization techniques, and/or data curation.
Experience with Python, C++, or other related programming languages, and deep learning frameworks such as PyTorch or TensorFlow.
Must obtain work authorization in the country of employment by the time of hire and maintain ongoing work authorization during employment.

Nice To Haves

Demonstrated research and software engineering experience through internships, industry or academic work, coding competitions, or widely used open-source contributions.
Experience with video generation, video synthesis, or multimodal fusion techniques.
Experience with video-language models and architectures relevant to video generation and post-training.
Experience with LLM/VLM agents, reward modeling, or evaluation of generative models.
A record of significant results demonstrated through grants, fellowships, patents, or publications at leading workshops, journals, or conferences such as NeurIPS, ICML, ICLR, CVPR, or ICCV.
Experience manipulating and analyzing complex, large-scale, high-dimensional data from varying sources.
Experience solving complex problems and evaluating alternative solutions, tradeoffs, and perspectives to determine a path forward.
Experience working and communicating cross-functionally in a collaborative, interdisciplinary team environment.
Intent to return to a degree program after completing the internship/co-op.

Responsibilities

Conduct fundamental and applied research in video generation, including generative models, video synthesis, and multimodal learning.
Develop and optimize post-training paradigms for large-scale video and multimodal models, including reinforcement learning from human and AI feedback and reward modeling.
Research self-improving and agentic generation systems, including multi-agent LLM/VLM pipelines and LLM/VLM-as-a-judge.
Collaborate with teams across Meta to build perceptual foundations for real-time embodied agents and conversational AI.
Contribute to the development of frontier models, such as Llama and large multimodal models, and advance video and media generation.
Brainstorm with research mentors, review literature, and evaluate existing solutions to challenging real-world research problems.
Develop novel solutions, implement prototypes, and run experiments using meaningful benchmarks and metrics.
Analyze experimental results and verify conclusions.
Draft and polish research reports and/or publications.
Present research outcomes to internal and/or external audiences.
Contribute research that can be applied to Meta product development.