crizo.ai
Apply now

AI doesn't speak your language yet.

Get paid to teach it.

Tested, trained, and calibrated. Applications open in 34 languages, including ten with no transcribed speech dataset in existence.

Become a Language ExpertRequest a sample dataset
  • 34languages
  • 10with no transcribed speech corpus in existence
  • 1,000+hours delivered

Open language roles

We work in languages most datasets skip entirely. Find yours.

Recording work needs only a phone. Transcription and annotation work needs a laptop.

Active

Running on live projects now.

Expanding

Onboarding experts now, ahead of project start.

How it works

Application to first paid work in 3–4 days.

3–4 days end to end

  1. 1

    Apply & assess

    ApplyAssessment

    A test in your language, not a generic form. A Bodo speaker is tested in Bodo.

  2. 2

    Train & calibrate

    TrainingCalibration

    Paid training on our annotation standards, then a calibration check against our quality bar.

  3. 3

    Feedback & deploy

    FeedbackDeployment

    Written feedback on your calibration, then assignment to a live project.

How you get paid

  • Paid per validated hour of accepted work.
  • Paid twice a month, on the 5th and the 20th.
  • Paid by UPI or direct bank transfer.
  • You will never be asked to pay anything, at any stage.

Who we're looking for

  • Native or near-native fluency in the language you apply for.
  • Formal education or professional work experience in that language.
  • At least 10 hours a week.
  • No prior AI or data-annotation experience needed. We train you.

From the people doing the work

  • The management team is highly supportive, approachable, and cooperative. I have received all my payments on time, directly into my bank account, without any delays.
    Rafiuddin Ali Ahmed Nagnoori

    Rafiuddin Ali Ahmed Nagnoori

    Kannada Language Expert

  • The tasks were clear, the instructions were easy to follow, and the support team was responsive whenever I had questions.
    Kanda Adi Venkata Satya Sai Sudha Rani

    Kanda Adi Venkata Satya Sai Sudha Rani

    Telugu Language Expert

  • I am very grateful for this opportunity. I have truly enjoyed the professional workflow and the clear, legitimate targets provided for each project.

    Tamil Language Expert

  • Every audio is like a mini detective case 😅. Learning a lot, improving every day, and enjoying the experience with a supportive team
    Ruthuvarshan V

    Ruthuvarshan V

    Malayalam Language Expert

  • The onboarding process was smooth, the instructions were clear, and the support team was responsive whenever I needed help. Thank you for providing a reliable opportunity to learn and grow.
    Orsu Anuhya

    Orsu Anuhya

    Telugu Language Expert

  • Working with this team has been a great experience. The support is prompt, the communication is clear. I appreciate the opportunity and look forward to contributing more in the future.

    Kannada Language Expert

  • The work environment is supportive, the communication is clear, and I've had the opportunity to learn and grow while managing transcription projects.
    Pushparaj P

    Pushparaj P

    Tamil Language Expert

Become a Language Expert

Know someone who speaks a language we're building?

The fastest way we find experts in rare languages is through the experts we already have. When someone you refer completes 5 hours of verified work, you get ₹2,000.

Refer someone

What we collect and deliver

Audio collection

  • Code-switched speech
  • Spontaneous conversation
  • Call-centre and phone-simulation dialogue
  • Scripted and read speech

Audio transcription

  • Code-switched verbatim transcription
  • Speaker diarization with millisecond-precision timestamps
  • Emotion and acoustic tagging
  • Internal WER / CER / DER tracking on every delivery

For AI labs

Depth in the Indic languages that don't have a commercial dataset yet — built through a managed pipeline, not ad hoc sourcing.

  • Managed data operations

    Recruit, train, validate, deliver. Run end-to-end in-house.

  • Human evaluation

    Every dataset passes expert human review, not automated QC alone.

  • Where the gap is

    India’s largest open speech corpora cover the 22 scheduled languages. The country speaks far more than 22. Ten of the languages we work in have no transcribed speech corpus at all, and the largest open collection that reaches beyond the scheduled list is under 10% transcribed.

  • Consent and rights

    Every recording is collected with documented contributor consent. Full commercial rights transfer on delivery. Contributor personal data is never included in delivered datasets.

Request a sample dataset

    Questions

    Do I have to pay anything to join?

    No. Not at any stage, for any reason. Crizo will never ask you for money, and anyone who does is not us.

    Can I do this on a phone?

    For recording work, yes — a smartphone is all you need. Transcription and annotation work needs a laptop. Every language card shows which type of work is open.

    How much internet do I need?

    Recording work is light on data. Transcription, annotation and model verification need a stable connection throughout the task.

    How and when do I get paid?

    Per validated hour of accepted work, paid twice a month on the 5th and the 20th, by UPI or bank transfer.

    How much time does this take each week?

    A minimum of 10 hours a week. You choose when.

    Can I do this alongside a job or studies?

    Yes, as long as you can commit 10 hours a week. Most of our experts do this alongside other work.

    How is my voice recording used, and who owns it?

    Recordings are collected with your written consent and used to train and evaluate AI speech systems. Crizo owns the recordings you produce for us. Your personal details are never included in anything we deliver to a client.

    Do I need prior experience?

    Not in AI or data work — we train you. You do need native or near-native fluency in the language you apply for, plus formal education or professional work experience in it.

    My language isn't listed. Can I still apply?

    Yes.

    Become a Language Expert

    Optional 30-second voice sample