What the work actually involves

You receive batches of Italian audio — recorded speech that may include regional accents, overlapping speakers, background noise, code-switching into English, and false starts — and produce transcripts that follow a client style guide down to the punctuation. The guide, not your instincts, decides whether you write disfluencies, how you tag inaudible spans, whether numbers are spelled out, and how speaker turns are marked. A second strand of the work is review: reading someone else's transcript against the audio and correcting completeness and consistency errors. A third is annotation — labelling segments for the attributes the client specifies, which might be speaker change, non-speech events, dialect markers, or audio quality.

The quiet part of the job is judgement about ambiguity. Most of the disagreement between transcribers concerns a two-second stretch where the speaker could plausibly have said one of two things, or where a Neapolitan or Sicilian pronunciation makes a word resolvable only by context. Strong contributors resolve what the guide lets them resolve, flag what it doesn't, and write a usable note rather than silently guessing — because a guessed word that reads fluently is more damaging to training data than an honest `[unintelligible]` tag.

What the screen looks for

micro1's process is AI-led and largely conversational, with a short asynchronous interview followed by practical assessment. Expect it to probe genuinely native-level Italian comprehension — not textbook Italian, but degraded audio, regional variation, and colloquial registers — plus evidence that you have worked to an external style guide rather than your own preferences. Concrete answers travel: the tool you used, the turnaround you held, the QA metric you were measured against, the type of content you handled. Prior AI work is explicitly not required and claiming it vaguely helps less than describing ordinary transcription discipline precisely.

Logistics

  • Fully remote contractor engagement; no fixed shifts stated, but batches carry deadlines tied to project milestones.
  • Async by default — coordination with project leads happens in writing, and clarification questions are expected rather than penalised.
  • You supply your own setup: reliable internet, decent headphones, a quiet place to listen closely for long stretches.
  • Volume on data projects fluctuates. Treat it as variable contract work that may ramp and taper rather than a steady weekly commitment.
  • Confidentiality obligations apply to all audio you handle.