What the work involves

You receive a set of prompts covering questions of French law — procedural points, questions of admissibility and delay, evidentiary rules, remedies, enforcement, appeal strategy. You run each prompt through two AI platforms, read both outputs side by side, and score them against a rubric supplied by the client. Every evaluation is accompanied by short written feedback in which you say, in plain terms, what the model got wrong: a mis-stated délai de prescription, a citation to a repealed article of the Code de procédure civile, advice that would be correct in Belgium or Québec but not before a French juridiction, or a confident answer to a question where the case law is genuinely divided.

This is comparative benchmarking, not content writing. The value you add is the reflex of a practising avocat who notices that an answer is structurally plausible and substantively unusable. Most reviewers find the reading is fast and the writing is the slow part — the rubric rewards feedback that is specific enough for an engineer who does not read French law to act on it.

What the platform screens for

  • Law-firm litigation experience — at least two years, three or more preferred, in litigation, disputes or contentious practice. In-house counsel qualify only if the move in-house happened within the last two years and was preceded by firm practice.
  • Qualification to practise in France as an avocat/e, with working command of French procedural law rather than general civil-law familiarity.
  • Native or fluent French and precise written expression; the screen probes whether you can state a legal error crisply rather than at length.
  • Rubric discipline — evidence that you can apply someone else's scoring criteria consistently instead of substituting your own judgment of what makes a good answer.

Logistics

Fully remote and asynchronous. The initial commitment is around 10 hours, typically completed over one to two weeks, with strong potential for further work if calibration holds. All work is covered by an NDA, which includes the identity of the platforms being compared and the prompt sets. Pay is observed at $150–160/hr for this listing; Mercor sets rates per engagement and they are not guaranteed to persist across later batches.