SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Cantina Labs is a social AI company building real-time generative models for interactive characters. You'll work on real-time video generation that brings characters to life—realistic humans, stylized characters, and non-human identities—with consistent appearance, expressive motion, and natural speech-to-movement synchronization.
In this role, you'll tackle interconnected research problems spanning streaming generation, fine-tuning, distillation, inference optimization, and data development. You'll have substantial freedom to propose research directions, investigate novel ideas, and shape priorities. The work balances quality, controllability, and efficiency, requiring judgment about which interventions (new training approaches, better data curation, evaluation frameworks, or implementation bottleneck fixes) are worth pursuing.
Key areas you'll explore:
- Real-time and streaming generation: develop models supporting responsive interaction while maintaining visual quality and consistency over time; investigate latency, continuity, and model behavior during extended generation.
- Character expression and motion: explore how appearance, facial expression, gesture, and body movement cohere into a coherent character; improve speech-to-motion relationships across identities and visual styles.
- Fine-tuning and post-training: adapt video models to new data, behaviors, and capabilities; investigate training objectives and conditioning methods through implementation and evaluation.
- Distillation and inference optimization: accelerate generation and improve efficiency across model-level and runtime approaches; measure effects on quality, latency, and memory.
- Data development: work directly with training examples; identify coverage gaps and failure modes; improve curation, sampling, and data quality.
- Evaluation: build comparisons that make progress visible and failures understandable; combine quantitative metrics with careful inspection and interactive testing.
- Research and collaboration: engage with relevant literature, reproduce and adapt useful ideas, share early experiments and findings, and help the team prioritize compute allocation.
You'll iterate quickly, share results early, and have freedom to explore new approaches.
Requirements:
- 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.
