Tech Stack

Description

You will build tools and infrastructure to automate research workflows for reward model development, enabling researchers to iterate faster and scale reward development across domains.

Requirements

• Prior research experience
• Strong Python skills
• Experience with ML workflows, data pipelines, and related infrastructure/tooling
• Comfortable working across stack from data pipelines to user-facing tooling
• Results-oriented with bias for flexibility and impact

Responsibilities

• Design and build infrastructure for reward signal iteration, including rubric development and robustness evaluation
• Develop automated systems for reward quality assessment and detection of reward hacks
• Create tooling to compare reward methodologies (preference models, rubrics, programmatic rewards)
• Build pipelines to reduce toil in reward development from dataset preparation to deployment
• Implement monitoring systems to track reward signal quality during training runs