Research Scientist: Robotics Foundation Models

Honda Research Institute USA; San Jose (CA); On-site; USD 120,000 - 180,000; Full time

Job summary

Honda Research Institute USA is recruiting a Research Scientist in Robotics Foundation Models to advance multimodal AI for dexterous robotic manipulation. The role integrates VLMs/VLAs with learning‑based methods, across simulation and real hardware environments.

You will collaborate with global R&D teams, publish results, and contribute to a portfolio of patents and prototypes, with a start date in mid‑2026. The position is based in San Jose, CA, on‑site.

Qualifications

• Ph.D. in robotics, mechanical engineering, electrical engineering, computer science, or a related field.
• Hands‑on experience applying VLMs, VLAs, or multimodal foundation models in robotics or embodied settings.
• Experience with model training, fine‑tuning, or architectural development of multimodal/foundation models.
• Familiarity with transformer‑based and multimodal models.
• Strong programming skills in Python or C++ and familiarity with ROS.
• Self‑motivated to advance the project and deliver on time.
• Strong track record of publications in robotics and machine learning venues.
• Strong written and oral communication skills.
• 1+ years of relevant work experience.

Responsibilities

• Research, design, adapt and apply multimodal foundation models for robotic systems to develop robust manipulation behaviors and task failure reasoning.
• Train, fine‑tune, and design vision‑language and multimodal foundation models for robotic dexterous manipulation.
• Research and develop machine learning algorithms such as reinforcement learning and imitation learning for dexterous manipulation.
• Implement and validate algorithms in simulation and on robot hardware.
• Analyze system performance and improve robustness and generalization.
• Deliver results in accordance with project timelines.
• Prepare written and oral technical reports and demonstrations.
• Contribute to portfolio of patents, academic publications, and prototypes to demonstrate research value.
• Collaborate with our teams of scientists and engineers in Honda’s regional and global R&D offices, as well as with our partners in academia and industry.

Skills

Python

C++

Reinforcement learning

Multimodal models

Transformer-based models

ROS

Education

Ph.D. in robotics / related field

Tools

ROS

Job description

Research Scientist: Robotics Foundation Models - Honda Research Institute USA

Honda Research Institute USA (HRI‑US) is seeking a passionate and self‑motivated research scientist specializing in Robotics Foundation Models to join our Robotics Division. This role focuses on developing robotic AI systems that combine multimodal foundation models (e.g., VLMs, VLAs) with model‑ and learning‑based methods to enable reliable and adaptable dexterous manipulation behaviors in real‑world environments.

San Jose, CA

Key Responsibilities

• Research, design, adapt and apply multimodal foundation models (e.g., VLMs, VLAs) for robotic systems to develop robust manipulation behaviors and task failure reasoning.
• Train, fine‑tune, and design vision‑language and multimodal foundation models for robotic dexterous manipulation.
• Research and develop machine learning algorithms such as reinforcement learning and imitation learning for contact‑rich dexterous manipulation.
• Implement and validate algorithms in simulation and on robot hardware.
• Analyze system performance and improve robustness and generalization.
• Deliver results in accordance with project timelines.
• Prepare written and oral technical reports and demonstrations.
• Contribute to portfolio of patents, academic publications, and prototypes to demonstrate research value.
• Collaborate with our teams of scientists and engineers in Honda’s regional and global R&D offices, as well as with our partners in academia and industry.

Minimum Qualifications

• Ph.D. in robotics, mechanical engineering, electrical engineering, computer science, or a related field
• Expert knowledge on learning‑based methods such as reinforcement learning, generative models, and inference.
• Hands‑on experience applying VLMs, VLAs, or multimodal foundation models in robotics or embodied settings.
• Experience with model training, fine‑tuning, or architectural development of multimodal/foundation models.
• Familiarity with transformer‑based and multimodal models.
• Solid knowledge of fundamentals of robotics, including kinematics, dynamics, control systems.
• Experience with planning, control, and system integration for robotic manipulation.
• Strong programming skills in Python or C++ and familiarity with ROS.
• Self‑motivated to advance the project and deliver on time.
• Strong track record of publications in robotics and machine learning venues.
• Strong written and oral communication skills.
• 1+ years of relevant work experience.

Bonus Qualifications

• Hands‑on experience on design, development, and fine‑tuning of foundation models for dexterous or multi‑fingered manipulation.
• Hands‑on experience with policy deployment and evaluation on real robot stacks.

Desired Start Date

7/20/2026