What the work involves

You are building the test set, not seeing clients. A typical block of work starts with authoring a counseling scenario from scratch: a presenting problem, enough clinical context to make the case ambiguous rather than textbook, and a turn where the counselor has to make a judgment call — a disclosure that edges toward duty-to-warn, a client minimizing substance use, an adolescent asking about confidentiality, a couple where one partner discloses IPV in a joint session. You then write the reference response and, critically, the reasoning: which modality or framework you're drawing on, what you assessed before responding, what you deliberately did not say and why.

The second half of the work is grading. You read model output against your rubric and score it for clinical accuracy, ethical handling, and risk management. The failure mode the labs care most about is the response that reads warmly — validating, reflective, well-worded — while skipping a risk assessment, colluding with avoidance, or giving advice that no licensed clinician would give unsupervised. Flagging that is easy; writing two or three sentences that name the exact clinical error, so an engineer with no counseling background can act on it, is the skill being paid for.

What the platform screens for

  • A verifiable, active, unencumbered LPC or LMHC license in a US state, plus two or more years of post-licensure practice. This is checked, not assumed.
  • Risk fluency under follow-up. Screeners tend to ask a safety question, then push: what changes your assessment, what do you document, what do you do when the client walks it back.
  • Named modalities, not vibes. Saying you use CBT is weaker than saying which intervention you'd sequence first and why for a specific presentation.
  • Written precision. Your grading notes are the deliverable. Vague criticism ("not therapeutic enough") fails; specific criticism ("reflected affect but never assessed means or intent after a passive ideation disclosure") passes.

Logistics

Fully remote and asynchronous — no live sessions, no caseload, no on-call. You set your hours each week against a minimum commitment of about ten. Work arrives as ongoing project batches, and pay is hourly via Stripe on a weekly cycle. The $65–100/hr band reflects rates observed on this platform for this role; actual rate depends on specialty depth and calibration performance, and is not guaranteed.