Research Scientist, Post-Training — Video Generation

Pika Research Scientist, Post-Training — Video Generation 22 Minutes AgoSaved In-Office Palo Alto, CA, USA 185K-400K Annually Junior 185K-400K Annually JuniorInformation TechnologyResearch Scientist responsible for RL-based post-training of large-scale video generation models. The role includes preference optimization, online RL, video reward model development, human and automated evaluation, reward-hacking safeguards, and secondary distillation of aligned models. Candidates need experience with generative modeling or post-training, reinforcement learning, diffusion or flow-matching models, PyTorch, and distributed training. Preferred experience includes visual-generation reward models, VLM-based judging, preference data collection, and video-specific temporal and physics failure modes.Top Skills: Diffusion ModelsFlow-Matching ModelsModel DistillationMulti-Node Distributed TrainingPreference OptimizationPyTorchReinforcement LearningVideo Reward ModelsVlm-As-Judge