Founding ML Evaluation Layer for Robotic Policy Validation

One Robot; San Francisco (CA); On-site; USD 140,000 - 210,000; Full time

Job summary

One Robot builds task-specific world models and an evaluation platform for robot manipulation policies. We train end-to-end policies and iterate rapidly, validating where policies fail and what data to collect to fix them.

Our team focuses on robust evaluation, VLM/LLM training, and data-driven improvements, working with real customer data and deployments. Based in San Francisco, we value ownership and fast iteration.

Qualifications

• Strong coding in Python and PyTorch.
• Track record in training VLMs or LLMs.
• Developed and shipped evals for VLMs or LLMs.

Responsibilities

• Train evaluation models to classify and verify policy behavior.
• Build confidence layers that translate model outputs into trustworthy signals for customers.
• Improve model grounding to reason about physical and spatial scenes.
• Build a self-improving evaluation layer to sharpen eval models with deployments and corrections.

Skills

Python

PyTorch

VLM training

LLM training

Eval development

Tools

Python tooling

Job description

One Robot builds task-specific world models and an evaluation platform for robot manipulation policies. We train end-to-end policies and iterate rapidly, validating where policies fail and what data to collect to fix them.

Our team focuses on robust evaluation, VLM/LLM training, and data-driven improvements, working with real customer data and deployments. Based in San Francisco, we value ownership and fast iteration.