Senior AI Trainer

What you'll do

Despite the title, this is generalist evaluation work rather than a specialist track. You test AI chat and search tools under standardised, repeatable conditions, then grade, rank and annotate their responses against detailed guidelines, with written feedback on why one answer beats another.

The second half is adversarial. You write and refine prompts that push the models into edge cases and unusual user intents, so the evaluation set covers more than the easy questions. Then you go back over ratings (yours and others') to catch places where the score and the rubric disagree, and fix them.

Findings go into the project's training tools as evaluation datasets, and you take part in quality reviews and in the discussions that settle ambiguous guideline cases across the team.

Who fits

• Someone with exceptional attention to detail when reading AI-generated content
• Clear written and spoken communication, including proactive updates when something blocks you
• Enough self-organisation to work alone and remotely against milestones
• An analytical mindset that goes looking for the cause of a disagreement instead of averaging it away

No degree and no minimum years of experience are stated.

Nice to have

• Prior RLHF, data labelling or other data-centric evaluation work
• Hands-on familiarity with AI products, prompt engineering, and where models tend to fail
• Experience writing rubrics or grading model outputs for accuracy, helpfulness and safety
• A background in ML data preparation or LLM-focused projects, which the ad marks as strongly preferred

The card lists video annotation and video annotation tools among the required skills, but the description itself is about chat and search evaluation. Expect the screening to ask about both, and do not assume video work is excluded.

What it pays

$14–36/hour, micro1's published band. The top is more than double the bottom and the ad does not say what decides placement, so plan around $14 until you have an offer. No compensation structure is published, so whether this is billed hourly or worked out per task is not stated.

Where it is open

The ad names Northern America and Europe as the location. That is a region rather than a country list, so check the screening questions for which countries are actually accepted.

Worth knowing

Good:

• 50 openings and posted within the last day, flagged high demand
• A low bar to entry: preferred experience is preferred, not required
• Prompt writing and rubric reconciliation are more varied than pure rating
• Useful as a first step into AI training work if you have none yet

Less good:

• A low floor for work titled "senior"
• Wide band with no published basis for placement
• The skills list and the description do not quite line up, so the actual task mix is uncertain
• No weekly hours, project length or start date stated
• Contractor engagement: no benefits and your own tax to manage

About this listing

Posted by micro1 on its own jobs portal and read on 24 September 2026, when it was one day old. The $14–36/hour band, the 50 openings and the Northern America and Europe location are its published text. The customer behind the project is not named. See micro1.