What the work involves

You'll spend most sessions in one of three modes. Authoring: producing reference presentations from a brief — a 12-minute conference talk on a given result, a poster layout for a mid-sized meeting, a figure panel for a grant proposal's significance section. Evaluating: scoring model-generated decks against rubrics that cover argument structure, figure legibility, slide density, audience calibration, and whether the claimed result is actually supported by what's on the slide. Refining: rewriting a model's output into something you'd genuinely show a study section or a departmental seminar, and writing the rationale for every change you made.

The rationale is the product. A model can already produce a slide with a title and three bullets; what it cannot do is know that a funder audience needs the aim restated on every section divider, that a busy results slide reads as hedging, or that a poster read at 1.5 metres has a different typographic floor than one read on a screen. Your written reasoning is what gets learned from, so a two-line note explaining why the axis label was wrong is worth more than a silently fixed deck.

What the platform screens for

Ethos runs an AI voice screen. Expect it to verify that your research activity is current — recent publications, recent talks, a lab or group you're actually attached to — and then to probe presentation judgment with follow-ups. Generic answers about "clear communication" get pushed on; the screen is looking for whether you can name a specific audience, a specific constraint, and a specific decision you made because of it. Slide craft is tested too: expect questions about how you'd handle a dense results figure, a strict time limit, or a non-specialist audience.

  • PhD with 3+ years of post-doctoral research activity, with a publication record
  • Demonstrable conference presentation history and experience presenting to funders or review panels
  • Expert PowerPoint skill — not just typing into templates, but layout, masters, figure handling, and export discipline

Logistics

Fully remote, fully asynchronous, no fixed hours. The stated range is 5–20 hours per week with room to take more, and tasks are generally claimed from a queue rather than scheduled. The $70/hour figure is the band observed for this listing; rates on these programmes vary by task type and can be revised, so treat it as reported rather than promised. Most contributors work in blocks of two to four hours, since deck tasks don't slice neatly into fifteen-minute gaps.