French Audio QA Specialist — Transcription & Diarization

About OpenTrain AI

OpenTrain AI is the hiring and contracting organization for this role. OpenTrain is the centralized platform where people start and grow careers in AI training and data labeling—work that helps shape how modern AI systems behave. Creating an OpenTrain account is free and gives you access to projects, a unified profile, and a way to build a durable freelance career in this fast-growing field.

Why AI Training Work Matters

AI training (also called data labeling or annotation) is the human layer behind every major AI system: people transcribe audio, label speakers, validate timestamps, and rate model outputs so models learn from high-quality examples. This work is often 100% remote, flexible, and accessible to contributors with language expertise and attention to detail. As a reviewer, you will be the final quality gate that keeps datasets precise, safe, and useful for multilingual AI research and products.

The Role

OpenTrain AI is recruiting a French Audio QA Specialist to serve as the final quality gate on a dataset of transcribed, multi-channel audio recordings. You will review technical audio fidelity, confirm human-validated transcriptions meet low error-rate targets, validate diarization and timestamps, and audit metadata and content for safety and privacy. This is a contract, part-time opportunity with a weekly commitment of 20+ hours.

What You'll Do

As the audio QA specialist you will perform detailed, technical reviews and provide clear feedback. Your work directly impacts dataset quality and how downstream multilingual, multi-speaker AI systems are evaluated.

• Evaluate multi-channel audio for technical fidelity: channel isolation, background noise, clipping, and gain issues.
• Review and verify human-validated transcriptions for exceptionally high accuracy and strict unnormalized verbatim rules.
• Validate turn-level and word-level timestamps and speaker identification, including in overlapping speech and complex multi-speaker scenarios.
• Read and validate JSON-formatted files to confirm timestamp logic, metadata tagging, and data integrity.
• Audit sessions for safety and privacy: detect PII, toxic or sensitive content, and confirm metadata correctness.
• Assign pass/fail or agree/disagree decisions and write detailed, actionable comments when annotator work is incorrect.

Requirements

You must meet the core qualifications below to be considered. These are substantive, role-specific requirements driven by the dataset quality needs.

• Native-level proficiency in French.
• Exceptional ear for audio fidelity and ability to detect subtle background noises, channel bleed, or clipping.
• Meticulous attention to detail for verifying word-level timestamps and strict verbatim transcription rules.
• Ability to assess complex multi-speaker dynamics including overlaps, false starts, and backchannels.
• Comfort reading and validating JSON-formatted data for accurate metadata tagging and timestamp logic.

Preferred Background

Candidates with the following backgrounds often excel in this role, though these are helpful rather than mandatory.

• Linguists or phonetics experts with deep understanding of spontaneous, unnormalized speech.
• Language teachers with native-level mastery and strict adherence to transcription guidelines.
• Professional transcriptionists experienced with rigorous timestamping and low word-error-rate targets.

Logistics & How to Apply

This is a contract, part-time position (CONTRACTOR, PART_TIME) requiring 20+ hours per week. The role is open worldwide and requires native French; no country restrictions are listed in this posting. The listing does not include a specific pay rate—compensation details will be shared during the application or onboarding process.

To apply, create a free OpenTrain account, complete your profile with language and audio QA experience, and submit your application for this project. Successful applicants will be asked to complete a short qualification task to demonstrate transcription accuracy, timestamp validation, and JSON-reading ability.