Research Scientist Post Training Video Generation
Pika is an active AI creative platform for generating and editing video, images, and audio, with a creator-facing app and an API offering.
Funding history
About Pika
Pika, operated by Mellis, Inc., provides a generative-AI platform through its websites, mobile apps, APIs, and third-party platforms. Its current product covers video, image, and audio creation/editing and offers access to first-party Pika and third-party models through an aggregated API platform.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will lead RL-based post-training for video diffusion and flow-matching models at multi-node scale. You will build video reward models, design preference-data workflows, train and validate learned judges, prevent reward hacking, run human and automated evaluations, and distill aligned models into efficient samplers.
Requirements
- 2+ years of hands-on research experience in post-training or generative modeling
- RL or preference-optimization experience on generative models with evidence of model improvement
- Knowledge of diffusion or flow-matching models
- PyTorch proficiency
- Multi-node distributed training experience
Responsibilities
- Run RL post-training for video diffusion and flow-matching models at multi-node scale
- Build video reward models and define evaluation dimensions
- Design and configure preference-data collection workflows
- Train and validate learned judges and safeguard against reward hacking
- Own post-training evaluation using human preference studies and automated metrics
- Distill RL-tuned models into efficient few-step samplers while preserving alignment gains
Benefits
- Equity
- Health benefits
- 401k matching
- Flexible onsite/remote hybrid work