Robotics Research Engineer: Vision-Language & World Models

Relari, inc.; San Francisco, Northern (CA, KY); Hybrid; USD 140,000 - 210,000; Full time

Job summary

Relari, inc. seeks a researcher to advance robotics foundation models, training pipelines, evaluation loops, and real-world robot deployment. You will build representations from multimodal data, train models, and use failures to refine ideas.

The role focuses on depth in machine learning or robot learning, applying Python-based tooling (PyTorch, JAX) to design experiments and measure progress on physical robots. Collaborative, hands-on environment values robust software for robotics experiments.

Qualifications

• Strong foundations in ML or robot learning.
• Experience building and evaluating learning systems in Python (PyTorch, JAX).
• Ability to turn open-ended questions into working experiments with clear conclusions.

Responsibilities

• Own pre-train, post-train, and evaluation pipelines of robotics foundation models.
• Develop vision-language-action models, video models, and world models.
• Investigate representations for transferring skills across robot embodiments.
• Build large-scale training, simulation, and real-world evaluation loops.
• Deploy policies on physical robots and diagnose failures across data, perception, learning, control.
• Improve tooling for data pipelines, teleoperation, retargeting, visualization, and runtime software.
• Read literature, reproduce ideas, and design focused experiments around results.

Skills

Machine learning foundations

Robot learning

Computer vision

Python

PyTorch

JAX

Tools

PyTorch

JAX

Python

Job description

Relari, inc. seeks a researcher to advance robotics foundation models, training pipelines, evaluation loops, and real-world robot deployment. You will build representations from multimodal data, train models, and use failures to refine ideas.

The role focuses on depth in machine learning or robot learning, applying Python-based tooling (PyTorch, JAX) to design experiments and measure progress on physical robots. Collaborative, hands-on environment values robust software for robotics experiments.