The work

Each task starts with you sourcing a publicly accessible Gujarati PDF in an assigned document type — a newspaper page, a textbook spread, an exam paper, a flyer, form, manual, menu or notice — that contains at least one multimodal element: an image, table, diagram or handwriting. You record where you got it, then map the page: bound every meaningful region, assign it a component type (document title, section heading, paragraph, list, table, figure, diagram, caption, formula, question, answer field), give it a reading-order index, and link child regions to their parent figure or table. Then you transcribe every text region exactly as it appears — in Gujarati script, not transliteration — including handwritten content, flagging regions where the source genuinely isn't legible. Page-level metadata (language, document type, source, page dimensions, flags for tables, formulas and handwriting) closes out the task.

The dataset is built deliberately around what document models handle worst: dense multi-column layouts, mixed-script pages, handwritten forms, tabular exam material. Nothing is generated by a parsing model — component identification, typing, reading order and transcription are all human-authored. As you build a track record, you move into reviewing other Gujarati experts' tasks end to end.

What the screen looks for

The platform screen is AI-led and probes three things. First, native Gujarati fluency with full command of the script — diacritics, conjunct forms, numeral forms, and the difference between correct orthography and what a keyboard happens to produce. Second, document experience: annotation, transcription, translation, localization, subtitling, proofreading, journalism, OCR post-editing or regional-language data review. Third, taxonomy discipline — whether you can apply a fixed set of component types consistently across hundreds of pages instead of improvising per document. Expect follow-ups that push on specific edge cases rather than accepting general claims.

Logistics

  • Fully remote and asynchronous; open only to contributors based in India, the United States, Canada or Western Europe.
  • Task-based work with no fixed shifts; observed pay on this listing is $13/hr, stated as observed and not guaranteed.
  • Accuracy is the primary metric, with handling time tracked alongside it. Tasks that fail second-expert review come back to you.
  • You need a computer with a working Gujarati input method and reliable access to public document sources.