Outlier is hiring worldwide — AI Trainers, Coding Experts & Writing EvaluatorsFreelance • Fully remote • $15–$60/hr • Work when you wantBrowse Outlier roles on MMagic.ai →
Outlier is hiring worldwide — AI Trainers, Coding Experts & Writing EvaluatorsFreelance • Fully remote • $15–$60/hr • Work when you wantBrowse Outlier roles on MMagic.ai →
R

Research Scientist, Video Generation

Featured
RunwayCompensation undisclosedRemote (Global)Posted 1w ago
PyTorchDiffusion ModelsComputer VisionGenerative AIDistributed TrainingTransformersCUDALatent Video SynthesisDeep LearningJAX

About the role

Join the team at the forefront of the generative video revolution to push the boundaries of what is possible with Gen-1, Gen-2, and beyond. As a Research Scientist, you will conduct foundational research to improve the temporal consistency, visual fidelity, and controllability of our large-scale video diffusion models. Your work will directly translate into creative tools used by millions of artists and filmmakers worldwide, shaping the future of storytelling through breakthroughs in latent video synthesis.

Responsibilities

- Design and train the next generation of video foundation models, focusing on long-form consistency and high-resolution synthesis. - Investigate novel architectures beyond standard Transformers and U-Nets to improve computational efficiency. - Develop advanced conditioning techniques including depth, motion vectors, and text-to-video alignment. - Implement and scale distributed training loops across hundreds of GPUs using internal infrastructure. - Collaborate with the engineering team to optimize inference speed and deployment of complex generative pipelines. - Stay current with the rapidly evolving SOTA and contribute back to the research community through high-impact publications. - Build internal tools and benchmarks to quantitatively evaluate video quality and temporal coherence.

Requirements

- PhD in Computer Science, Machine Learning, or a related field, or equivalent practical research experience. - Proven track record of publications in top-tier venues such as CVPR, ICCV, SIGGRAPH, or NeurIPS. - Deep expertise in generative modeling, specifically diffusion models, autoregressive models, or GANs. - Experience training large-scale models (LLMs or Vision) on massive datasets using distributed systems. - Strong proficiency in PyTorch and experience with low-level optimization for hardware accelerators. - Ability to thrive in a fast-paced, iterative environment and transition research into production-ready code. - Strong mathematical background in stochastic calculus, linear algebra, and optimization.
Disclaimer: MMagic.ai connects talented people with AI companies around the world. While we work hard to feature quality opportunities, we don't independently verify employers, candidates, salaries, or hiring outcomes. We encourage you to research each opportunity and company before applying or making an offer.

Similar roles