Member of Technical Staff, TLM: Research Scientist

What we're looking for

As we scale our research ambitions, this role is for a scientist who doesn't just execute on known problems but defines new ones. This TLM role sits at the intersection of deep technical research and scientific leadership: you'll own a long-horizon agenda, mentor a small group, and help shape what this field becomes.

What you'll do

• Define and drive original research into the foundational open problems of world modeling: temporal coherence, causal reasoning, long-horizon prediction, and user-intent alignment.
• Move beyond diffusion and transformer paradigms, design novel architectures, learning objectives, and training frameworks that reframe what generative interactive video can do.
• Build and mentor a small, high-output research group, guiding work from hypothesis through experimental validation to prototype demonstration.
• Translate foundational breakthroughs into stable prototypes in close collaboration with engineering. Not incremental improvements, but genuinely new capabilities.
• Publish at top-tier venues (NeurIPS, ICML, ICLR) and actively engage the broader research community to help Odyssey shape the direction of the field.

Who you are

• A staff-level or senior ML researcher with a strong publication record in generative modeling, video synthesis, multimodal learning, or adjacent areas.
• Someone who works from first principles: you define the problem before reaching for a method, and you build minimal systems to prove what's possible before scaling.
• Motivated by understanding: you care about why models work, not just that they do.
• Experienced leading or mentoring a small research group, with a track record of taking projects from open question to publication or deployment.
• Comfortable in a focused, autonomous lab environment where deep thinking and tight collaboration are the norm.
• Excited to help establish the scientific foundation of a new medium, one that will define how AI perceives and interacts through pixels over the next decade.

Our staff is expected to be in office at least three times a week. A lot of our progress comes from working through problems together, and we want you in the room for that.