Research Analyst

About the Role

We are developing an evaluation benchmark for advanced AI browsing agents. Your main responsibility will be to devise research challenges that remain unsolvable by leading AI systems, even when granted unrestricted web access and multiple attempts. This role focuses on investigative research rather than subject-matter expertise or content creation.

You will begin with a verifiable fact and reverse-engineer a research question that makes discovering that fact exceedingly difficult. Your work will be validated through a comprehensive, traceable evidence trail.

Deliverables

• Craft a natural-language research question with a concise, stable, and objectively verifiable answer.
• Provide multiple independently verifiable clues covering diverse fact types such as dates, people, locations, organizations, works, events, records, and quantities, each with clear constraints.
• Compile a validation report outlining the specific searches performed and the outcomes achieved.

Required Qualifications

• Proven expertise in conducting open-web research including sourcing primary documents and navigating various databases, archives, registries, and PDF files.
• Exactness in citation practices, referencing particular pages, tables, and sections beyond homepage links.
• Ability to research unfamiliar topics effectively from the ground up.
• Native or near-native proficiency in written English.
• Strong aptitude for maintaining detailed structured documentation, as the evidence trail forms the core of the task.
• Hands-on experience in evaluating large language models (LLMs), red-teaming exercises, or designing benchmarks.
• Professional background in at least one of these fields: reference librarianship, archival research, investigative journalism, professional fact-checking, open-source intelligence (OSINT), due diligence, know-your-customer (KYC) processes, patent or legal discovery research, genealogy, or competitive puzzle and quiz construction.

Desirable Skills

• Knowledge of JSON and structured data formats for delivering research outputs.