Founding Research Scientist

present of AI; San Francisco (CA); On-site; USD 150,000 - 230,000; Full time

Job summary

Verne Robotics seeks a senior researcher to translate robot learning ideas into trained models and verify them on a real robot arm.

You will work with hardware and engineers handling control and deployment, publishing first-author results and driving rigorous experiments with clear baselines.

Qualifications

• Hands-on research role taking a robot learning idea from hypothesis to trained model.
• Strong Python and PyTorch skills with a track record of first-author papers.
• Comfort working with hardware and collaborating with control/deployment engineers.

Responsibilities

• Build AI models from video and demonstrations to physical manipulation skills.
• Train, test, and evaluate on real robot arms with clear baselines.
• Run careful experiments and collaborate with hardware and software engineers.

Skills

Python

PyTorch

Robot learning

Experiment design

Hardware familiarity

Publish first-author papers

Tools

Robot arm hardware

Control systems

Job description

AI & Data · San Francisco · FullTime · Senior level

This is a hands-on research role for someone who can take a robot learning idea from first hypothesis to a trained model and then test it on a real robot arm. A strong candidate has a record of first-author papers at top venues, solid Python and PyTorch skills, and the habit of running careful experiments with clear baselines. The posting also expects comfort working with hardware and with engineers who handle control and deployment.

How it fits the AI era

The core of the job is building AI itself: models that learn from video and robot demonstrations and turn them into physical skills. AI tools can speed up coding, experiment tracking and data analysis, but deciding which questions matter, reading why a robot failed, and judging what data is missing stay with people. Real robots expose problems that benchmarks hide, so careful human judgment about results carries a lot of weight.

How it builds AI experience

The role offers direct experience training and fine-tuning large models that combine video, sensor data and actions, then checking them on physical arms. Since the company releases an open-weight model, someone preparing could study it, try fine-tuning on a small manipulation task, and write up the results. Hands-on time with a low-cost robot arm and teleoperated demonstrations would also show readiness.

Worth showing

• First-author robot learning papers
• PyTorch model training and debugging
• Rigorous experiments and ablations
• Real robot manipulation results
• World action or diffusion policies

This is our own read, written from the employer's posting and the company's Spotlight. The employer's posting is the source of truth for the role.

About Verne Robotics

Verne Robotics builds AI models that let robot arms learn new skills in hours, per the company.