Pika
Research Scientist, Post-Training — Video Generation
US remote · On-site · Full-time · Posted May 16, 2026
- Experience
- 2+ years hands-on research experience in post-training or generative modeling.
About the job
Own post-training and evaluation of video generation models, including RL alignment, reward modeling, and secondary work distilling models into efficient samplers.
What you'll do
- Run multi-node RL post-training for video diffusion or flow-matching models.
- Build and validate video reward models, including preference data workflows and safeguards against reward hacking.
- Lead post-training evaluation with human preference studies and automated metrics; distill RL-tuned models as a secondary focus.
What they ask for
- Hands-on research experience in post-training or generative modeling.
- RL or preference-optimization experience on generative models, with evidence of model improvement.
- Strong grounding in diffusion or flow-matching models, PyTorch, and multi-node distributed training.
Nice to have
- Experience developing visual-generation reward models, VLM-as-judge, or large-scale preference data collection.
- Distillation expertise and familiarity with video-specific failure modes.
Tools
PyTorch
AI focus
RL alignment and post-training for video generation models; development of video reward models and evaluation; secondary focus on distilling RL-tuned models.
Model-written summary · the original posting is authoritative