Staff Scientist, Interactive World Models & Robotics

Odyssey; Palo Alto (CA); On-site; USD 180,000 - 240,000; Full time

Benefits offered by this job

In‑person collaboration

Job summary

Odyssey is an AI lab pursuing general world models for continuous interaction with the real world. We seek researchers who can advance diffusion-based world models and robot-policy training, with hands-on work across data, models, and deployment to real hardware.

The role values experimental rigor and a track record of delivering end-to-end ML projects, including sim-to-real work and scalable training on modern GPUs. In-person collaboration in the global hubs is common.

Qualifications

• A PhD (or equivalent research experience) plus 2+ years of relevant research/engineering experience, or 4+ years software engineering with 2+ years ML work.
• Hands-on experience with diffusion models, world models, model-based RL, or VLA policies.
• Track record of owning projects end to end.
• Comfort across the ML pipeline from data to real-robot deployment; sim-to-real is a plus.
• Proficiency with PyTorch (or TF/JAX).

Responsibilities

• Learn what makes interactive world models tick, including data, diffusion backbones, action conditioning and policies.
• Implement state-of-the-art diffusion and robot-learning algorithms, design losses and RL fine-tuning.
• Run ablations across WM and policy spanning architectures and data mixes.
• Build data pipelines, WM fine-tuning, and policy deployment onto Robots-as-a-Service and partner systems.
• Exploit latest GPU features to increase training and inference efficiency.
• Take ownership of the full ML stack used by Odyssey researchers and product engineers.

Skills

PyTorch

Diffusion models

World models

Model-based RL

Robotics data pipelines

ML research

TF/JAX

Education

PhD or equivalent

Job description

Odyssey is an AI lab pursuing general world models for continuous interaction with the real world. We seek researchers who can advance diffusion-based world models and robot-policy training, with hands-on work across data, models, and deployment to real hardware.

The role values experimental rigor and a track record of delivering end-to-end ML projects, including sim-to-real work and scalable training on modern GPUs. In-person collaboration in the global hubs is common.