R

Redwood Research

Nonprofit AI safety and security research

Conceptual Reasoning Capabilities Researcher

AI ResearchPosted on Nov 9
Full-timeBerkeley, CA, USA$350k - $850k /yrIn-person

About the Job

Requirements

  • Assess the logical validity of arguments and weigh the strength of fuzzy, hard-to-formalize arguments.
  • Be willing to spend substantial time carefully evaluating many LLM-written arguments or manually writing rubrics with hundreds of items.
  • Have a deep interest in either AI risk or acausal interactions.

Responsibilities

  • Create datasets for evaluating and improving AI at conceptual reasoning.
  • Depending on skills and experience, build scaffolds for models, run training experiments, develop project ideas, communicate research and theory of impact, manage other researchers, or manage relationships with AI labs and other stakeholders.

Desired Qualifications

  • Develop novel ideas and arguments.
  • Have experience in a conceptual domain.
  • Have experience interacting with LLMs, including using coding agents, building scaffolds, or engineering prompts.
  • Have experience training LLMs.
  • Have management experience.

About the Company

Redwood Research

Nonprofit AI safety and security research

Redwood Research conducts AI safety and security research focused on preventing advanced AI systems from acting against human and institutional interests. It studies how misalignment could occur as AI capabilities rise and develops methodologies to manage these risks while preserving the benefits of AI. The work emphasizes understanding risks, evaluating safety strategies, and building practical approaches for alignment, verification, and governance. Its goal is to reduce the likelihood and impact of dangerous AI behavior and to guide responsible development and deployment of powerful AI systems.

Company size
11-50
Company stage
Not listed
Total funding
Not listed
Headquarters
Berkeley, CA
Founded
2021

Recently Posted Jobs

Explore more opportunities at Redwood Research

See more jobs