Audio Data Quality Specialist - JT AI Labs (M/F/D)

About this role

As one of the first members of JT AI Labs, you will own the quality of the audio datasets we collect, from the first applicant sample to the dataset the client accepts: ● Review audio by ear and with tools: noise, echo, clipping, distortion, silence, cross-talk and leakage between tracks, narrowband or Bluetooth input, and wrong sample rates or formats. ● Check that recordings meet each project’s specification: duration, speaker profile, language and accent, spontaneity of speech, and absence of personal data the expert shouldn’t have shared.

● Review applicants’ audio samples in our review console and help decide who is accepted. Monitor speaker quotas (age, gender, region, accent, hours per speaker) against client targets. ● Write and maintain QC guidelines, rejection reasons and the issue taxonomy. Turn them into clear recording instructions and training material for experts. ● Run and extend our QC scripts on session exports. Work with the data engineer to decide which checks get automated and to tune their thresholds against your human judgement.

● Design and run the human sample review that complements automated QC. Calibrate reviewers and measure their agreement so decisions stay consistent across projects. ● Track and report quality metrics: QC rejection rate per project and per expert, recurring issues, client acceptance rate. Propose improvements to the platform and the recording setup. ● You have a degree in Audio Engineering, Sound Production, Phonetics or Linguistics, Data Science, or a related field, or equivalent professional experience.

● You have 2+ years of experience in audio quality control, speech data QA, transcription or annotation QA, or audio post-production. ● You have a trained ear and can reliably hear subtle problems: background noise, compression artefacts, clipping, echo, mic handling, channel imbalance. ● You understand the technical basics of digital audio: sample rates, bit depth, codecs, lossless versus lossy formats, mono and multi-track recordings.

● You are comfortable with audio tools such as Audacity, Praat, iZotope RX or a DAW, and can read a waveform and a spectrogram. ● You can write and adapt small scripts in Python or Bash to batch-check files, parse metadata or compute statistics (ffmpeg, sox, soundfile, pandas or similar). You don’t need to be a software engineer, but you automate anything you do twice. ● You understand what makes speech data useful for training ASR or TTS models, or you are keen to learn.