Prompt/QA Engineers for AI Study

I'm running a formal validation study of AI test-coverage methods and need two experienced prompt/QA engineers for a short, well-defined engagement. You'd be one of two independent specialists; you won't interact with each other. The work (4 tasks, ~5–8 hours total, spread over ~2 weeks): 1. Failure-scenario enumeration (~4 × 1 hr, async). For each of four real AI prompts (a support router, a triage assigner, and similar), you get the prompt and one hour: list the scenarios where you'd expect it to fail. AI tools are allowed and expected — use whatever assistants you normally would, with one restriction on which model family (disclosed at kickoff), and a short log of what tools you used. Deliverable: a failure-scenario list per prompt + one 15-minute debrief call. 2. Expert labeling session (~1 hr, live on a video call). You'll tag a set of test cases against provided definitions, sharing your screen, no AI assistance for this one task — the whole point is your unaided judgment. Paid at the same rate; scheduled at your convenience. 3. Failure clustering (~1 hr, async). Given a set of failing AI outputs (source hidden), group them into distinct failure modes. AI assistance fine here. 4. On-call tiebreaker pool (~0–1 hr). Occasionally adjudicate items where crowd raters split. Same no-AI rule as task 2; done in the labeling app, low volume. You're a fit if you: have shipped or QA'd production LLM features (prompt iteration, eval design, red-teaming, or test authoring); can articulate *why* a prompt fails, not just that it does; are comfortable being timed and logged (it's a study — process is data); and can commit to availability windows across two specific weeks (dates at kickoff). Logistics: $45–75/hr depending on experience; ~5 hrs guaranteed, up to ~8 if the tiebreaker pool activates. Async except the labeling session and one debrief call. NDA-free — everything you see is external/public material, and your deliverables are anonymized in any write-up. I'll ask two short screening questions before hiring (below). To apply, answer the two questions (a paragraph each, no essays) below.