What the work actually is
You write tax problems that a strong model should plausibly get wrong, then you show your work. A typical unit might be a multi-state individual return with a part-year residency wrinkle, an S-corp reasonable-compensation question, a partnership basis and 704(b) allocation trace, or an estimated-tax safe-harbor calculation where the prior-year AGI crosses the 150k threshold. Each item needs a defensible answer, a citation trail to the Code, regs, or IRS guidance, and enough distractor structure that a plausible-sounding wrong answer is genuinely available. The second half of the job is grading: reading model output, deciding whether it reached the right number by the right reasoning, and writing feedback that names the exact step where it went off — not "incorrect," but "applied the 2023 standard deduction to a 2024 fact pattern and then compounded it in the QBI limitation."
What the screen is looking for
AfterQuery's intake is AI-led and follow-up heavy. It probes whether your tax knowledge is load-bearing or memorized: expect to be pushed one layer past your first answer, asked where a rule has an exception, and asked what changes if a date or a filing status moves. Reviewers care that you can distinguish a settled answer from a position that depends on facts, elections, or an unsettled area — and that you say so rather than manufacturing certainty. Writing quality matters concretely, because your feedback text is the training signal.
Logistics
- Fully remote and asynchronous; no standing meetings, deadlines are per-batch.
- Contract work, hourly, invoiced against logged time; volume fluctuates with project pipeline.
- Most contributors run 10–20 hours a week, though some batches allow more.
- Observed pay band is $75–150/hr, with the top of the range typically going to CPAs or EAs with corporate, partnership, or international depth. Rates are as reported and not guaranteed.
- A CPA, EA, or MST is preferred rather than required; demonstrated compliance experience carries real weight.