Mercor is seeking a Japanese Audio Generalist Evaluator Expert to contribute to a high-impact audio AI research project with a leading research lab. In this role, you will work on transcription, annotation, and evaluation tasks that help train and benchmark advanced language models. This is a short-term, structured engagement ideal for candidates with strong academic or analytical backgrounds who are fluent in Japanese and English and enjoy translating complex audio and visual information into precise, well-structured text.
Job Responsibilities
Transcribe and Optimise Audio & Video
- Listen to, analyse, and transcribe audio and video content in Japanese, following detailed constraints and instructions.
- Produce high-quality written outputs in Japanese, with supporting work in English when required.
- Ensure clarity, accuracy, and strict adherence to formatting and stylistic guidelines.
- Capture nuances such as tone, intent, formal vs. informal register, honorific language, regional dialects, and spoken Japanese variations where relevant.
Define and Document Evaluation Standards
- Establish clear expectations for correct and high-quality responses in general consumer audio contexts.
- Develop detailed evaluation rubrics and grading guidelines in Japanese and English.
- Document standards to ensure consistency across reviewers and model evaluations.
- Identify linguistic nuances, grammatical complexities, cultural references, colloquialisms, and edge cases specific to Japanese.
Conduct Model Testing and Grading
- Run prompts through language models and assess generated outputs.
- Evaluate responses against predefined criteria for accuracy, completeness, fluency, and instructional clarity.
- Provide structured feedback to improve model performance in Japanese audio tasks.
Support Benchmarking and Quality Assurance
- Participate in QA and review cycles to ensure tasks, rubrics, and outputs meet Mercor’s quality bar.
- Maintain consistency and reliability before datasets are integrated into official benchmarks.
- Collaborate with project leads to resolve ambiguities and improve task design.
Minimum Qualifications
- Strong writing, editing, and critical thinking skills.
- Ability to work independently, manage time effectively, and meet deadlines.
- Native or near-native fluency in Japanese (spoken and written) and professional fluency in English.
- Strong familiarity with spoken Japanese, regional dialects, honorific systems (keigo), and contemporary language usage.
- Ability to accurately transcribe and analyse Japanese audio content across general consumer contexts.
- Available to commit 10–20 hours per week.
Preferred Qualifications
- College students or recent graduates.
- Background in linguistics, humanities, social sciences, journalism, translation/localization, or technical disciplines.
- Prior experience with transcription, annotation, localisation, evaluation, or research workflows in Japanese.
- Familiarity with regional dialects and variations across Japan.
- Interest in AI, language models, or applied research environments.
Application & Onboarding Process
- Complete a short AI-led interview (approximately 15 minutes).
- If selected, you will be onboarded and invited to begin project work.