About the role
Overview
Are you passionate about languages, writing, and quality evaluation? We are launching an exciting language evaluation project and are looking for Japanese-speaking experts to join our team.
What you will do:
Identify realistic search-related failures where an AI model provides inadequate responses. Create search queries, test them, document what went wrong, and write instructions and evaluation criteria for correct answers.
Common issues: local relevance, source quality, outdated information, language mismatches, incomplete answers, and region-specific inapplicability. Testing multiple queries before finding failures is expected.
This is a practical, execution-focused role. We value hands-on performance and consistency over advanced academic training. Candidates who prefer theoretical/research work may not find this role a good fit.
Project Details:
- Start Date: ASAP
- Duration: Ongoing Project.
- Commitment: 20/hours week.
- Job Type: Freelance contract.
- Location: Japan, Remote
- Rate: $35-57.75 per hour.
Requirements:
- Research skills — ability to find and verify information from multiple sources; comfort navigating different websites, databases, and documentation
- Critical thinking — Can spot contradictions between sources, identify outdated information, and recognize when something doesn't apply to a specific locale/context
- Attention to detail — Can catch nuances in language, local variations (prices, regulations, requirements), and edge cases
- Problem-solving mindset — Comfortable iterating and testing when things don't work; doesn't get frustrated by trial-and-error
- Domain knowledge — Ideally someone with background in research, quality assurance, compliance, or similar work that requires precision and source verification
- Writing clarity — Can articulate instructions and rubrics clearly so others understand the logic behind their reasoning
- Locale knowledge — Deep familiarity with their specific locale (local regulations, business practices, cultural nuances) is valuable for spotting region-specific failures