What the work involves

AfterQuery collects speech data that frontier labs use to train and evaluate models on Indian languages and Indian-accented English. Day to day you'll work through prompt batches: reading scripted lines, holding unscripted conversational turns on assigned topics, and producing code-switched recordings that reflect how you actually move between your language and English rather than an artificially clean version of either. Each prompt comes with detailed specifications — target register, speaking rate, whether disfluencies are wanted or not — and following them precisely matters more than sounding polished.

The second half of the work is review. You'll listen to recordings (yours and other contributors') against their transcripts and correct errors: dropped words, wrong script or transliteration choices, misheard proper nouns, boundaries placed in the wrong spot on a code-switch. The hardest category is audio that sounds entirely natural but doesn't match what's written down — those files pass a casual listen and quietly poison a training set, so flagging them is a core deliverable.

What the platform screens for

  • Verifiable language depth. Which Indian language, which regional variety, where you acquired it, and whether your English is genuinely professional-register — not conversational-only.
  • Recording setup. Screens ask concretely about your microphone, your room, and the background noise you cannot control. Phone-speaker audio in a reverberant room does not pass.
  • Instruction discipline. Whether you can follow a prompt spec exactly, and what you do when a prompt is ambiguous or would produce unnatural speech.
  • Transcript judgment. Whether you can articulate what counts as an error versus an acceptable variant in transliteration, spelling, and code-switch notation.

Logistics

Fully remote and asynchronous — no scheduled calls, no fixed shifts. You set your own hours week to week, with a floor of about 10 hours. Pay is weekly via Stripe. The observed band is $20–$40/hr, with placement typically reflecting language rarity, audio quality, and whether you take on review work as well as recording; nothing here is a guaranteed rate. Project work is ongoing rather than a single fixed batch.