About the job
The Video Intelligence team is an applied AI research team within the Facebook pillar. This role is expected to develop advanced video generation and understanding foundation models, enabling innovative AI-driven video creation experiences and enhancing our ability to comprehend video content. The team is responsible for building State-of-the-art GenAI technology to empower video generation and understanding.
Responsibilities
Build a variety of multimodal foundation models such as text-to-video generative models, image-to-video generative models, video understanding models, unified native video generative models
Design core foundation model architectures and progressive pre-train
Post-train foundation models using techniques such as Supervised Fine-Tuning (SFT), Reinforcement Learning from Human Feedback (RLHF), Direct Preference Optimization (DPO), and Low-Rank Adaptation (LoRA)
Conduct research to develop SOTA GenAI models for the Facebook family of apps
Collaborate with colleagues from the infrastructure and product teams on launching models
Qualifications
Minimum
Bachelor's degree in Computer Science, Computer Engineering, relevant technical field, or equivalent practical experience
PhD in Computer Science, Machine Learning, or a relevant technical field
1+ year of industry experience training multimodal, computer vision, LLM or related AI/ML models
Experience owning and/or driving complex technical projects from end-to-end
Publications at peer-reviewed conferences (e.g. ICLR, NeurIPS, ICML, KDD, CVPR, ICCV, ACL)
Programming experience in Python and hands-on experience with frameworks such as PyTorch
Must obtain work authorization in the country of employment at the time of hire, and maintain ongoing work authorization during employment
Preferred
First-authored publications at peer-reviewed conferences (e.g. ICLR, NeurIPS, ICML, KDD, CVPR, ICCV, ACL)
Experience collaborating in cross-functional teams, including product, engineering, and research
Experience building text-to-video generative models, image-to-video generative models, video understanding models, and/or unified native video generative models