Dexterous Robotics ML Research Engineer

Bracket Bot Inc.; San Francisco (CA); On-site; USD 120,000 - 180,000; Full time

Job summary

Bracket Bot Inc. in San Francisco is seeking a Robotics Research Engineer to develop software and machine learning for dexterous manipulation.

You will turn research ideas into robot capabilities through experiments, model development, integration, and testing on physical hardware. This hands-on, onsite role involves collaborating with robotics, software, and hardware engineers to make manipulation behaviors reliable beyond a single demonstration.

Qualifications

• Strong programming skills and a foundation in robotics and machine learning.
• Experience implementing and evaluating models or robotics algorithms.
• Ability to design experiments and reason clearly about results.
• Comfort debugging software on physical robots.
• Clear communication and ownership from experiment through implementation.

Responsibilities

• Develop and evaluate software and machine learning approaches for dexterous robotic manipulation.
• Build data collection and evaluation workflows to measure manipulation performance and understand failures.
• Implement research ideas, run experiments, and compare approaches.
• Integrate learned models and manipulation behaviors into robot software.
• Test on physical robots, debug failures, and improve reliability.
• Document results and collaborate with software and hardware engineers to turn research into working capabilities.

Skills

Programming skills

Robotics foundations

Machine learning

Experiment design

Communication & ownership

Job description

Bracket Bot Inc. in San Francisco is seeking a Robotics Research Engineer to develop software and machine learning for dexterous manipulation.

You will turn research ideas into robot capabilities through experiments, model development, integration, and testing on physical hardware. This hands-on, onsite role involves collaborating with robotics, software, and hardware engineers to make manipulation behaviors reliable beyond a single demonstration.