Pietro Romeo built more than a job board for AI training work
Pietro Romeo, Founder & Product Owner of aitrainer.work, started the platform after finding AI training job sites hard to use. It combines job listings with learning resources, an Academy and a talent pool. The conversation looks at how user research, manual candidate matching and a search setback shaped its direction.
Start by talking to the people you want to help and identifying the questions they repeatedly ask. Keep listening after launch, and build useful answers to those questions rather than relying on generic ideas or large volumes of thin content.
What’s the short story of how your business started, and what makes it genuinely different from the other options your customers have?
I’m a UX researcher. In December 2025 I applied for UX research roles on one of the platforms that pay experts to help train AI models. I did the interviews, and while I waited to hear back I compared the other platforms in the space. The sites listing their jobs all looked alike: the same template with a different logo, and copy that made little sense. I launched aitrainer.work on December 31, 2025 to build the site I had wanted while I was waiting.
What sets it apart is that it’s built from user research. Changes start with what people tell us in feedback and bug reports. When a learner asked for flashcards in one Academy module, we added them within a day, then to the other modules. We also list fewer platforms on purpose. We check each one by hand, and we leave out platforms that pay in their own crypto token or that workers report badly. Each pay figure says whether the platform stated it or we estimated it.
How did you win your first customers, and where does most of your new business come from today?
Our users are people looking for AI training work, and the site is free for them. Google indexed the site within days of launch, and the first users came from search. In February a Google update dropped many of our pages. Traffic held up because Bing and AI chatbots like ChatGPT started sending people to us, which is a little ironic for a site about training AI.
Today new users come from search engines, AI chatbots and writers who cite our data. The other half of the work is matching. When a partner platform needs people for a specific role, I go through our Talent Pool and invite the candidates whose background fits. Since launch we have referred more than 7,000 candidates, and more than 450 have been hired.
What’s one decision that changed the direction of your business, and one mistake that taught you the most?
The decision was to stop being only a job board. People in this field apply to many platforms and often wait months before they get a project, so a list of openings doesn’t get them hired. In April I added a free Academy that prepares people for the screening tests, and a Talent Pool where I match candidates to roles by hand. That changed what we measure. The number that counts now is hires, and we publish it.
The mistake was launching thousands of pages that said little more than "here’s a job we found." Google cut many of them in its February update. Now each page carries data only we have, such as the pay platforms state for that kind of role and how many people we’ve seen hired.
What has changed in your industry or in what your customers expect over the last couple of years, and how have you adapted?
The biggest change is what AI labs want. They still hire people to rate and write answers, but now they are short of data from real businesses, such as internal documents, team chat and code, and they pay companies to license it. Most companies don’t know their data has value or who would buy it. So we launched a spin-off, Dataconto (dataconto.com), a deal desk that matches companies that have data to sell with the AI labs and data buyers looking for it. Twelve buyers work with it today.
For aitrainer.work, the change has been competition. Dozens of new job sites appeared in our space, and one copied ours down to the Academy module names. Our answer was to double down on the user experience. We talk to users, change the site based on what they tell us, and rewrite any page that reads like a machine wrote it.
If you were advising someone starting out in your industry today, what would you tell them to do first, and what would you tell them to ignore?
Do the research before you write any code. Talk to the people you want to help, watch how they look for work today, and write down the questions they keep asking. For aitrainer.work, those questions were what a platform pays, whether it’s legit and what the screening test is like, and most of what we built answers one of them. Keep asking after launch too. Users will tell you what to build next if you give them a way to.
Ignore the idea that AI can come up with your product for you. Ask a chatbot what to build and you get the same suggestions everyone else gets, which is how this space ended up with dozens of near-identical job sites. Treat AI as nothing more than a tool, not as a strategic business partner.
Thank you to Pietro Romeo and the team at aitrainer.work for sharing what actually worked with Leaders Perception readers.
Want to tell the story behind your business?
Leaders Perception publishes interviews with business owners and leaders every week. Five questions, about ten minutes, no calls. It is free and every completed interview gets published.