About

I'm Luke Lashley, an AI engineer. Part of my day job is reviewing what AI models actually did, step by step, so the work these platforms pay for isn't far from what I already do.

I have a computer science degree from the University of Georgia, and I'm finishing Georgia Tech's online master's in computer science at the end of 2026. Outside work I write open-source software, mostly for Home Assistant: I wrote its Roborock, Anova, Snoo and Harbor integrations, I help maintain Oral-B, and I'm the main contributor to python-roborock, the library behind the Roborock integration. Roborock alone shows up in more than 50,000 Home Assistant installs that share analytics. All of that is public on my GitHub.

I started paid AI training work in June 2026 and do about 20 hours a week across several platforms. Since then it's paid me nearly $30,000, most of it from DataAnnotation and micro1. I made this site to write down what happens between an advertised rate and an actual payment: applications, unpaid screenings, waiting for projects, and getting paid. Some of it went well. Some of it, like a long Outlier screening that paid $0, I'd skip next time.

How I write these reviews

Every number on this site is labeled as either something I was paid or something a listing advertised. The screenshots come from my own dashboards, with names, IDs and client project names hidden. When a platform changes how it pays or I get a new result, I update the review and its date.

My background is programming, so most of what I can see firsthand is coding and technical work. If you're coming from another field, the listed rates in each review and the job finder will tell you more than my own pay can.

Get in touch

Email me at admin@paidtotrainai.com. You can also find me on LinkedIn and GitHub, and the site has a community at r/paidtotrainai.

How this site makes money

Some links here are referral links: if you sign up through one and end up doing paid work, the platform may pay me. It costs you nothing, and the pay numbers and complaints stay in either way. Details are on the disclosure page.