Reviews and open roles
Someone has to teach these models.
Open right now
All 878 →Frontier AI models learn from people who already know a subject — lawyers, physicists, radiologists, machinists, translators. That work is hiring right now, and almost nobody will tell you plainly what it pays.
Platforms you sign up to once
These don't post separate roles. You make one profile, pass one screening, and they match you to projects as work comes in.
SuperAnnotate's network of subject-matter experts who train and test AI models. Medicine, law, finance, STEM and language roles.
Fits
Specialists with a degree or hands-on expertise who can wait for a project. Passing screening doesn't come with a start date.
Pay
SME Careers' own job page lists ceilings from $15/hr (language roles) to $120/hr (US legal rubric design), with $100/hr for US healthcare, finance and STEM rubric roles (264 open roles, Oct 16, 2026). Your rate is set per project.
Handshake's AI training program. Projects for students, graduates and professionals across many fields.
Fits
Graduates and professionals who want a large pool of projects. After signing up, accept the invite, then search the role title you came for.
Pay
Up to $25–$150/hr ceilings, published per role, per our review (checked Sep 12, 2026).
An expert network running AI training campaigns, with a rate published on every listing and no guaranteed volume.
Fits
Senior professionals with a named-firm background who want a published rate before applying.
Pay
$50–$150/hr advertised, per campaign, per our review (checked Sep 12, 2026).
What this work pays
$75/hr
Median rate across every role on this site. Half of them pay more, and the middle half fall between $53 and $103 an hour.
See what's openMedian listing midpoint $75/hr. Half the priced listings have midpoints between $53 and $103/hr.
Solid bars hold the middle half of the midpoints; outlined bars are the tails. These are rates the board advertises, not what anyone takes home.
Median listing midpoint $75/hr. Half the priced listings have midpoints between $53 and $103/hr.
Solid bars hold the middle half of the midpoints; outlined bars are the tails. These are rates the board advertises, not what anyone takes home.
Open roles by specialty
Pick your field and see three of its open roles, and the rate it pays.
Choose a specialty
The 3 most recently posted of 88 open roles in Sciences & Research. We haven't ranked these for quality or vetted them, and a rate a platform advertises isn't what anyone takes home.
Member of Technical Staff, Model Training (applied research)
micro1 · Sciences & Research · Posted October 27, 2026
$7–$8/hrApply →Member of Technical Staff, Model Training (infra & platform)
micro1 · Sciences & Research · Posted October 27, 2026
$7–$8/hrApply →Member of Technical Staff, Model Training (post-training research)
micro1 · Sciences & Research · Posted October 27, 2026
$7–$8/hrApply →
Also on the board, too small or too far down the list for a button here: Finance (57), AI & Machine Learning (47), Arts & Design (32), Robotics (19), Data Analysis (18), Energy & Power (15), Sales & Marketing (14), Cybersecurity (8), and 8 smaller ones. They're all on the full board.
The reviews
Every platform we cover
What each one is, who it hires, how you get in, and what people who have worked there say — including the parts that would talk you out of applying.
DataAnnotation
Published pay
No rate published
Fast entry, real hourly rates, and a steady stream of worker reports of accounts closed without notice.
Read the review →Ethos
Published pay
$70/hr median
An expert network running AI training campaigns at $50 to $150 an hour, with no guaranteed volume.
Read the review →Handshake AI
Published pay
up to $80/hr median
Rate ceilings published up to $150/hr, and a documented pattern of suspensions leaving finished work unpaid.
Read the review →Mercor
Published pay
$80/hr median
High published rates for credentialed professionals, reliable weekly pay, projects that end without notice.
Read the review →micro1
Published pay
$68/hr median
Pass a live AI interview, then wait to be matched. Published bands run $15 to $200+ an hour.
Read the review →SME Careers
Published pay
No rate published
A real expert network run by SuperAnnotate, with documented weekly pay through Deel.
Read the review →Turing
Published pay
No rate published
Real frontier-lab projects behind a long unpaid screening pipeline, and almost no published pay rates.
Read the review →
Why this site exists
Annotation Insider is written by people who have done this work. The information around it is bad: blanket pay ranges with no source, review sites quoting each other, and platforms that describe a rate without describing the hours around it.
Every open role, in one place
878 roles across 4 platforms, updated daily, filterable by specialty. No other site collects them.
Reviews that name the downsides
What each platform is, who it hires, how you get in, and what people who work there say about it — including the parts the platform would rather you skipped.
Matched to your background
Paste your resume and see which roles name skills you already have. It runs on your device and nothing is uploaded.
Tools that are actually free
A pay calculator that counts the unpaid hours, and trackers for applications, hours and tax set-aside.
The shape of an application to this kind of platform. Each review sets out that platform’s own, which differ more than you would expect — one of them runs no interview at all.
Step 1
Apply to a role
A short form, usually with a resume or an expertise profile attached.
Step 2
AI interview
Recorded and scored where a platform runs one; elsewhere an unpaid assessment does the same job.
Step 3
Decision, then a wait
Review windows vary, and a rejection often arrives with no reason and no appeal.
Step 4
Identity checks and paperwork
A photo ID check, a contractor agreement, tax forms and a payout account.
Step 5
Matching, then a first task
Getting through the gate is not the same as being given work, and that wait is rarely published.
Which platform fits you?
A few quick questions, including your domains of expertise. We point you at a platform and the open roles that match what you tell us — which is not the same as knowing a platform will take you.
Question 1 of 4
How many hours a week can you realistically commit?
Pick one — it moves you on.