Expert Talent Pool, Evals and Benchmarks (Associate)

About the Role

Join a specialist talent pool supporting human-data projects that evaluate AI models. You will help design and run assessments that measure model capabilities and contribute to AI safety and capability research.

What You'll Do

• Write questions and tasks that isolate and test specific model capabilities.

• Develop rubrics that define strong, partial, and poor answers.

• Create reference answers and edge cases to challenge AI models.

• Grade model outputs against rubrics and explain scoring decisions.

• Analyze failed answers, group error patterns, and document findings.

• Compare multiple models using identical prompts.

• Validate benchmarks and flag compromised or ineffective questions.

• Participate in small user studies evaluating ratings or interactions with model outputs.

What We're Looking For

• Zero to five years of professional experience.

• Experience in QA or test engineering, assessment design, psychometrics, UX research, data analysis, ML evaluation, research support, or relevant subject-matter work involving question writing.

• Experience writing test cases, exam questions, rubrics, or evaluation tasks, and reviewing work consistently against written criteria.

• Comfort using spreadsheets or Python to organize and analyze results, and identifying and documenting error patterns.

• Experience with evaluation tools such as Inspect, lm-eval-harness, HELM, or promptfoo is useful.

• Clear written communication and the ability to compare AI model outputs or participate in user and rating studies.

Compensation & Benefits

Compensation is $30 to $60 per hour, depending on project, field, and experience. Payment is for hours worked.

Location

Fully remote. Projects typically last from one week to several months.