Research Scientist - Embodied AI at Confidential Berkeley, CA

Research Scientist – Embodied AI

In this role you will work on-site in our Berkeley, CA office.

• Architect robotic foundation models and lead the design of large‑scale robot learning frameworks tailored for industrial manipulation and autonomous planning.
• Build full‑stack training pipelines for robotics applications integrating multi‑modal inputs and ensuring reproducibility.
• Advance data collection strategies to support model training and validation.
• Bridge the sim-to-real gap for embodied AI systems utilizing simulation frameworks.
• Create scalable inference loops that combine perception with real‑time motion control and on‑robot deployment.
• Work alongside cross‑functional teams to turn research prototypes into deployable robotics systems that are robust and safe.
• Present and communicate ideas and project results to internal and external project partners and at conferences.

• Ph.D. or M.Sc. in Computer Science, Electrical Engineering, Robotics, or similar from an accredited university.
• A proven track record of contributions at top‑tier robotics and AI venues such as IROS, ICRA, RSS, CoRL, etc.
• Professional‑level fluency in C++ and Python.
• Proof experience in building and deploying machine learning models on robotic systems, including training, evaluation, and integration.
• Deep technical knowledge of modern model architectures, specifically Transformers, VLAs/VLMs, and Diffusion Models, with experience in large‑scale training.
• Hands‑on experience implementing, debugging, and deploying software on physical robotic platforms.
• Solid understanding of embedded systems, hardware/software integration, and OS fundamentals.
• Proficiency in English, both written and verbal.
• Legally authorized to work in the United States without corporate sponsorship now or in the future.

• Expertise in 3D scene understanding, multimodal grounding, and sensor fusion.
• Familiarity with simulation environments such as Isaac Sim, Mujoco, or Gazebo, and demonstrated experience with sim‑to‑real transfer strategies.
• Working knowledge of robotics middleware (ROS/ROS2) and integration of machine learning components into real‑time robotic stacks.
• Experience in building or adapting foundation models for embodied tasks (RFMs, VLMs/VLAs, multimodal diffusion).
• Experience with cloud computing environments.

We value your unique identity and perspective. We are fully committed to providing equitable opportunities and building a workplace that reflects the diversity of society, while ensuring that we attract the best talent based on qualifications, skills, and experiences.