Real-Time Video AI Research Scientist

Cantina Labs; United States; On-site; USD 180,000 - 240,000; Full time

Benefits offered by this job

Competitive salary

Equity

Healthcare coverage

PTO 42 days

Parental leave

401(k)

Job summary

Cantina Labs is hiring in the United States for a role developing real-time video generation for interactive characters, spanning realistic humans, stylized figures, and non-human identities. You’ll ensure characters speak, react, and express themselves with coherent appearance and motion throughout interactions.

The role covers training and fine-tuning video models, optimizing inference, evaluating data quality, and collaborating across research and engineering to push the boundaries of

Qualifications

• Experience training or fine-tuning video or related generative models.
• Ability to learn unfamiliar methods quickly and turn ideas into useful experiments.
• Good experimental judgment, ownership, and clear communication.

Responsibilities

• Develop real-time and streaming generation for interactive characters while maintaining visual quality and continuity.
• Explore character expression and motion, and improve speech-to-motion relationships across identities and styles.
• Pursue fine-tuning, post-training, distillation, and inference optimization to improve performance and efficiency.
• Investigate data curation, sampling, and coverage to improve model behavior.
• Design evaluations that reveal progress and failures; analyze generated video and user interactions.
• Collaborate with researchers and engineers, reproduce ideas, and share early demos and findings.
• Identify which interventions are worth pursuing and propose directions for future work.

Skills

Video models

Fine-tuning

Experimentation

Communication

Tools

Python

PyTorch

Job description

Cantina Labs is hiring in the United States for a role developing real-time video generation for interactive characters, spanning realistic humans, stylized figures, and non-human identities. You’ll ensure characters speak, react, and express themselves with coherent appearance and motion throughout interactions.

The role covers training and fine-tuning video models, optimizing inference, evaluating data quality, and collaborating across research and engineering to push the boundaries of