What the work actually involves

You write prompts a model should plausibly fail, then judge whether it failed and why. In practice that means a mix of authoring — adversarial questions on voice-leading, modal mixture, tuning systems, a specific composer's revisions, mic placement on a drum kit, the lineage of a regional genre — and evaluation, where you read model output and mark it against factual accuracy, reasoning quality, completeness, and nuance. The last one is where music gets difficult: a response can be factually clean and still wrong in spirit about, say, what makes a Coltrane substitution function, or flatten a living tradition into a Wikipedia summary. You are expected to say so, in writing, with a citation.

Much of the value you add is in catching confident nonsense. Models invent catalogue numbers, misattribute works, date recordings wrong, blur genre chronology, and reason fluently from a false premise. Your written rationale is the deliverable as much as the score is — researchers use it to trace where the model's knowledge thins out. Expect to maintain consistency against a rubric across hundreds of items, and to document edge cases rather than quietly resolving them yourself.

What the screen looks for

Turing's minimum bar is a Master's degree in any field (music-related disciplines preferred) plus 3+ years of professional work in music — education, performance, composition, production, journalism, research, content creation, or audio engineering. The assessment tests whether you can hold detail under pressure: specific repertoire, specific theory, specific gear or recording practice, and whether you can explain why an answer is wrong instead of asserting that it is. Prior LLM, prompt engineering, or annotation experience is a plus, not a gate. Written English carries real weight here, since feedback prose is the product.

Logistics

  • Contractor assignment, 8 weeks, no medical or paid leave
  • 40 hours per week — this is a full-time commitment, not evenings work
  • Remote, with at least 4 hours of daily overlap with PST
  • Pay band undisclosed in the listing; confirm rate and any extension terms before you commit
  • Application runs through a take-home assessment, then internal review