QA Specialist – Audio Annotation & Diarization (Korean) | Contractor | Fully Remote | 15-Week Engagement
Turing is seeking a detail-oriented QA Specialist with native Korean proficiency to serve as the final quality checkpoint on an evaluation-grade, multilingual audio annotation dataset. You'll review multi-channel recordings, human-validated transcriptions, and diarization outputs to ensure they meet rigorous accuracy standards for frontier AI development.
About Turing
Based in San Francisco, Turing is the world's leading research accelerator for frontier AI labs and a trusted partner for global enterprises deploying advanced AI systems. Turing delivers high-quality training data, advanced pipelines, and top-tier AI expertise across coding, reasoning, STEM, multilinguality, multimodality, and agents.
Role Overview
This project involves building a highly accurate dataset of transcribed, multi-channel audio recordings used to evaluate multilingual, multi-speaker AI systems. Contributors record unscripted group conversations, which are then transcribed and diarized — and your role as QA Specialist is to validate and approve the final output end-to-end.
Key Responsibilities
- Audio Quality Assurance: Evaluate multi-channel recordings for technical fidelity, verifying channel isolation, appropriate recording environments, and absence of background noise, clipping, or low gain.
- Transcription & Diarization Review: Audit human-validated transcripts for strict verbatim accuracy, preserving natural speech elements such as overlaps, interruptions, and false starts, while meeting low Word Error Rate (WER) targets.
- Timestamp & Speaker Validation: Confirm precision of turn-level and word-level timestamps and speaker identification, including in complex overlapping dialogue scenarios. Review underlying JSON-formatted data for accurate metadata tagging and timestamp logic.
- Metadata & Content Review: Verify demographic markers, contextual domains, and conversational tags. Enforce safety and privacy standards by auditing for PII, toxic, or sensitive content.
- Reporting & Feedback: Assign clear pass/fail or agree/disagree statuses and provide detailed, actionable feedback when disagreeing with annotator work.
Requirements
- Native proficiency in Korean
- Exceptional ear for audio fidelity — able to detect subtle noise, channel bleed, or clipping
- Meticulous attention to detail for word-level timestamps and verbatim transcription
- Ability to assess complex multi-speaker conversational dynamics
- Comfort reading and validating JSON-formatted annotation data
Ideal Backgrounds
- Linguists / Phonetics Experts: Deep knowledge of natural speech patterns, overlaps, false starts, and backchannels
- Language Teachers: Native-level mastery with strict adherence to verbatim transcription, including disfluencies and filler words
- Professional Transcriptionists: Strong audio fidelity instincts and rigorous timestamping experience
Engagement Details
- Type: Contractor / Freelancer (no medical or paid leave benefits)
- Duration: 15 weeks
- Commitment: Minimum 40 hours/week, at least 4 hours/day with 4 hours of overlap with PST
- Environment: Fully remote
Perks
- Work remotely on cutting-edge AI projects with leading LLM companies
- Potential for contract extension based on performance and project needs
