What the work involves

You are not answering IP questions for a chatbot — you are constructing the problem sets that measure whether a model can act like senior IP counsel. A typical task starts with a scenario you design from scratch: a continuation strategy for a crowded semiconductor portfolio, a §103 rejection with a questionable motivation-to-combine, an EPO Article 123(2) added-matter problem, a freedom-to-operate review before a product launch, or a cross-border license with field-of-use carve-outs and improvement clauses. You then draft the reference output a competent practitioner would produce — an office action response, claim chart, patentability opinion, litigation memorandum, or license term sheet — and write a rubric that specifies exactly where a weaker answer goes wrong.

The rubric is the hard part. Generic checklists reward bar-exam recall; useful rubrics penalize the specific failure modes of confident models: claim construction untethered from the intrinsic record, prior art analysis that ignores what the reference actually enables, procedural advice that misses a deadline or an available petition route, indemnity language that reads fine but shifts risk the wrong direction.

What the screen looks for

  • Verifiable practice history: portfolios you prosecuted, matters you owned, negotiations you led — not exposure by proximity.
  • Depth under follow-up. Expect the AI interviewer to push on a case you describe until it reaches procedure-level detail (RCE vs. appeal reasoning, divisional timing under Rule 36 EPC, IDS obligations, PCT national-phase mechanics).
  • Credential specificity: bar admission plus USPTO registration number, or European Patent Attorney qualification / equivalent.
  • Evaluation judgment — whether you can articulate what separates a defensible answer from a merely fluent one.

Logistics

Fully remote, asynchronous, contractor engagement. Contributors commonly work 10–20 hours per week against task batches with their own deadlines rather than fixed shifts; some engagements scale up during active project windows. The observed pay band for this listing is $90–100/hr, set by track and depth of qualification — treat it as observed, not guaranteed. Tooling familiarity (PatSnap, Innography, PatFT/Patent Public Search, Anaqua, CPA Global, claim-charting workflows) helps you build realistic scenarios but is not itself the credential being tested.