Research Scientist, Frontier Risk Evaluations

Written by the T&S Network from Scale AI's posting.

Frontier AI risk evaluation work inside Scale Labs, a team that bridges AI research and policymakers on AI risks and capabilities. The role designs evaluation measures, harnesses and datasets that measure the dangerous capabilities of frontier AI models and agents.

• AI misuse and jailbreaks

What you'd do

• Build test harnesses that probe AI models and agents for hazardous abilities like exploiting security flaws and giving meaningful help toward chemical, biological, radiological or nuclear weapons.
• Build evaluation measures and datasets for measuring risks posed by frontier AI systems.
• Work with government agencies and other labs to scope and design evaluations for advanced AI risks.
• Put evaluation methods into the open and write technical reports aimed at policymakers.

What it asks for

• A published research record in machine learning, especially generative AI.
• At least three years addressing sophisticated ML problems in research or product development.
• Practical experience building and instrumenting ML pipelines and writing evaluation harnesses.
• Ability to turn research literature ideas into working prototypes quickly.
• Strong written and verbal communication for cross-functional teams.
• Experience crafting evaluations and benchmarks, or data science work on LLM technologies (preferred).
• Experience stress-testing AI systems through adversarial attacks or red-teaming (preferred).

Terms in this role

Red teaming.

Testing a system, often an AI model, by deliberately trying to make it fail or misbehave, so problems are found and fixed first.

Our best summary of the posting, and it can be wrong. The employer's posting is what counts: read it in full on Scale AInorth_east before you apply. If we got something wrong, report this listing.