About the role
We’re looking for Research Engineers to contribute to our ambitious research agenda spanning multimodal foundational models, interactive world modeling, and real-time generation. This role is highly collaborative and will touch many aspects of our broader research efforts, taking a full-stack approach across pretraining, SFT/RL post-training techniques, evaluation, and bringing models to production.
What you’ll do
- Run experiments to teach world models new behaviors — action following, scene manipulation, camera control, and beyond
- Develop and test new data strategies, architectural variants, and training techniques
- Design evaluations that measure model capabilities, and use them to drive model improvement
- Take features from research prototype to production, collaborating with product and creative teams to address application-specific gaps
- Contribute across the entire stack (data pipelines, modeling, production inference) to move projects forward
What you’ll need
- 4+ years of experience in machine learning research or engineering
- Familiarity with the architecture, training, and inference of large-scale multimodal generative models
- Experience building robust data pipelines for pretraining and post-training
- Comfort working across the full research stack: data, training, evaluation, and deployment
- Proficiency with at least one ML framework (e.g. PyTorch, JAX) and experience with distributed training at scale
- Ability to context-switch quickly and drive projects forward in a fast-moving, ambiguous environment
- Excitement about building AI that simulates the world