Multimodal Robotics AI Researcher for Dexterous Manipulation

Moonfire; San Francisco (CA); On-site; USD 140,000 - 210,000; Full time

Benefits offered by this job

Relocation sponsorship

Enhanced pension plan

Enhanced holiday & paid leave

401(k) with company contributions

Health, dental & vision coverage

Open PTO policy

Job summary

Flexion is hiring a Research Scientist/Engineer to advance generalizable robotic manipulation using video demonstrations, wearables, and multimodal data. You will own end-to-end experimentation, from capture hardware design to model deployment on real hardware.

You will shape scalable data pipelines, develop 3D reconstruction and simulation environments, and push the frontier of foundation models for dexterous manipulation in real-world settings.

Qualifications

• PhD in Computer Science or related field.
• Experience with vision, video, and other multimodal models.
• Experience with multimodal generative modeling, training, and inference.
• Experience with reinforcement learning and imitation learning.

Responsibilities

• Develop scalable human data integration pipelines using video demonstrations and wearable capture systems into foundation model and RL workflows.
• Build automated pipelines for 3D object and environment reconstruction from capture data.
• Create simulation-to-real environments leveraging reconstructed assets to train dexterous manipulation policies.
• Design, train, and evaluate robot foundation models translating multi-sensor data into robust control.
• Develop algorithms to bridge human embodiment to high-DOF dexterous hands and humanoid platforms.

Skills

PhD in CS or related field

Vision & multimodal models

Multimodal generative modeling

Reinforcement learning

Imitation learning

Education

PhD in Computer Science or related field

Tools

Simulation environments

Job description

Flexion is hiring a Research Scientist/Engineer to advance generalizable robotic manipulation using video demonstrations, wearables, and multimodal data. You will own end-to-end experimentation, from capture hardware design to model deployment on real hardware.

You will shape scalable data pipelines, develop 3D reconstruction and simulation environments, and push the frontier of foundation models for dexterous manipulation in real-world settings.