Research Scientist, Post-Training — Video Generation
Core
Refining video generation models using RL alignment and building robust video reward models for real-time creative platforms.
Role type
Staff/Lead Research Scientist (RL Post-Training & Generative Modeling)
Builds
RL-aligned video diffusion/flow-matching models and video reward models
Domain
Generative AI, Video Generation, Reinforcement Learning
Deliverable
production ML models
Required skills
RL post-training, preference optimization, generative modeling, diffusion models, flow-matching models, PyTorch, multi-node distributed training
Preferred skills
video reward model development, VLM-as-judge, large-scale preference data collection, model distillation, video-specific failure mode analysis
Technologies
PyTorch, diffusion models, flow-matching models
Responsibilities
Run RL post-training (preference optimization, online RL) for video models at multi-node scale; Build and validate video reward models; Own post-training evaluation including human preference studies; Distill RL-tuned models to efficient samplers
Seniority
Staff/Lead, hands-on IC
