What the work involves
You'll listen to pairs and batches of AI-generated Hindi songs and decide which output is better, then explain why in terms an engineer would recognize. Sessions typically cover musicality and arrangement, prompt adherence, vocal quality and pronunciation, lyrical coherence in Hindi, and production/mix quality — clipping, muddy low-mids, unnatural sibilance, phase issues, dynamics. The second half of the work is annotation: tagging genre and sub-genre (filmi, ghazal, bhangra, indie Hindi, devotional, and beyond), marking song structure, and describing vocal and instrumental characteristics against a fixed taxonomy the lab supplies.
Good work here is specific. "Sample B is better" is worth little; "Sample B holds the tanpura drone in tune through the antara, while A drifts about 20 cents flat and the vocal doubling smears the consonants" is what the lab is training against. Expect rubric updates mid-project and occasional calibration rounds where your scores are compared against other raters.
What the platform screens for
Mercor runs a roughly 25-minute AI-led conversational interview after your resume and application form. It probes verifiable production credits, how long you've worked in Hindi-language music specifically, and whether you can hold up under follow-up questions about mixing decisions, theory, and lyric writing. Vague claims tend to collapse quickly — have specific projects, roles, and technical choices ready to describe. Bilingual fluency matters in both directions: you'll read English instructions and rubrics while judging Hindi lyrics and diction.
Logistics
- Fully remote and largely asynchronous; you claim tasks from a queue
- Part-time, 10+ hours per week, up to a 6-month engagement
- Immediate start; onboarding details typically follow within a few days of the interview
- Observed pay band is $9.63–17.33/hr, varying with assessed depth and task type — not guaranteed
- Good headphones or monitors and a quiet listening environment are effectively required
