What the work actually involves

You will be handed tax documents — often scanned, sometimes poor quality, sometimes form types you have not seen before — and asked to produce a structured representation of them. Day to day that means drawing tight bounding boxes around every labelled field, assigning the correct data type (string, date, currency, checkbox, enumeration), and then tagging the relationships that make a tax form a tax form: which line items feed a subtotal, which schedule attaches to which line on the 1040, which box on a W-2 corresponds to which entry downstream. You will also be asked to write down field metadata — format constraints, validation rules, what a valid value looks like — so a model can be scored against it later. Reviewing your own exports in JSON or YAML to catch mislabels before they ship is part of the job, not an extra.

What the screen is looking for

Two things, and most candidates only have one. The first is genuine familiarity with tax form structure — not just recognising a 1099-NEC, but knowing what depends on what, and being able to open an unfamiliar state or corporate schedule with the instructions beside it and work out the field logic yourself. The second is annotation discipline: consistent box boundaries, correct handling of multi-line and checkbox fields, sensible decisions when a form is rotated, cut off, or has handwriting bleeding across a box edge. Expect follow-up questions that push on specific forms and specific edge cases rather than accepting a general claim of experience.

Logistics

  • Fully remote, freelance, contracted through Turing.
  • Up to 40 hours per week; roughly a one-month engagement with possible extension based on performance and project need.
  • Largely async annotation work, with schema and guideline discussions as needed; throughput and consistency matter more than fixed hours.
  • Pay is not disclosed in the listing — treat any figure you see quoted elsewhere as observed, not guaranteed, and confirm rate and currency before you start.

A background in accounting, bookkeeping or tax preparation is listed as nice-to-have and is the most common route in. Prior work on document AI training data is the other.