Web Research Benchmark Designer

The work

You will design research problems that test whether AI browsing agents can find and verify information on the open web. The work is investigative: you will start with verifiable facts and build questions that remain difficult even when an agent can search repeatedly.

You will create structured research outputs with clear evidence trails. Each answer must be objective, stable, and independently checkable.

• Write natural-language research questions with short, objectively verifiable answers.
• Develop clues involving dates, people, places, organizations, works, events, records, and quantities.
• Research government and institutional databases, archives, registries, and PDF documents.
• Record obvious searches and the results they return.
• Cite exact pages, tables, and sections so each answer has a complete validation trail.
• Deliver organized research data; familiarity with JSON is helpful.

What it pays and takes

This is a worldwide, part-time contract role for an intermediate-level contributor. The project requires at least 20 hours per week.

• Pay: $30 per task.
• Schedule: 20+ hours per week.
• Location: Open worldwide.
• Language: Fluent written English.
• Education or experience: A master's degree or more than three years of relevant experience.
• Research: Demonstrated open-web research ability and confidence investigating unfamiliar subjects from scratch.
• Documentation: Strong sourcing precision and comfort creating detailed evidence trails.
• Relevant AI experience: Experience with LLM evaluation, red-teaming, or benchmark construction.
• Helpful backgrounds: Reference librarianship, archival research, investigative journalism, fact-checking, OSINT, due diligence, KYC, patent or prior-art searching, legal discovery, genealogy, competitive quizzing, or puzzle-hunt construction.

How it works

Apply on OpenTrain with your resume, then complete the application on the hiring site.

About AI training work

AI training is the human work behind systems that learn from examples, including writing and evaluating questions, checking answers, and reviewing model behavior. People with strong research and verification skills are valuable because they help make AI evaluations accurate, repeatable, and grounded in reliable sources.