AI Robotics Research Scientist - World Models & VLM/VLA

Socket.dev; Milpitas (CA); On-site; USD 140,000 - 200,000; Full time

Benefits offered by this job

Stock options

Health insurance

Visa sponsorship

Team events

Job summary

RoboForce in Milpitas, CA is seeking researchers to advance AI-powered robot control through vision-language models and world-models. You will design and deploy VLMs/VLAs, train world models for long-horizon planning, and integrate multimodal data for natural human-robot communication.

The role emphasizes producing production-grade solutions with in-office collaboration five days a week and a strong track record in foundation models, PyTorch/JAX, and robotics.

Qualifications

• PhD in ML/Robotics or related field, or Master’s with 4+ years experience.
• Proficiency in Python and DL frameworks (PyTorch, JAX).
• Expertise in large foundation models (VLM, VLA) and world-model architectures.

Responsibilities

• Design and deploy vision-language models (VLM/VLA) for contextual understanding and robot action policies.
• Develop and train world models for action-conditioned prediction and long-horizon planning.
• Research approaches to improve world model fidelity with multimodal inputs (vision, language, proprioception).
• Develop foundation models with spatial reasoning for precise robotic actions.
• Integrate multimodal data for natural human-robot communication.
• Optimize and deploy models as production-grade solutions on RoboForce platforms.
• Requires 5 days/week in-office collaboration with teams.

Skills

Python

PyTorch

JAX

Transformers

World models

VLM/VLA

Education

PhD or Master with 4+ years relevant experience

Tools

TensorRT

CUDA

Job description

RoboForce in Milpitas, CA is seeking researchers to advance AI-powered robot control through vision-language models and world-models. You will design and deploy VLMs/VLAs, train world models for long-horizon planning, and integrate multimodal data for natural human-robot communication.

The role emphasizes producing production-grade solutions with in-office collaboration five days a week and a strong track record in foundation models, PyTorch/JAX, and robotics.