Utkarsh Amitabh says he definitely wasn’t in the market for a new job in January 2025, when data labeling startup micro1 approached him about joining its network of human experts who help companies train artificial intelligence models.
The U.K.-based, 34-year-old entrepreneur already had a busy schedule as an author, university lecturer, founder and CEO of global mentorship and careers platform Network Capital, and student working toward a Ph.D. at the University of Oxford’s Saïd Business School. He also had a newborn at home, he tells CNBC Make It.
Ultimately, Amitabh agreed to take on the added role, admitting that “intellectual curiosity drew me in,” he says. The prospect of training enterprise AI models felt like a perfect fit with his own background in “business strategy, financial modeling and tech,” he adds.
Indeed, micro1 says it recruits experts with deep knowledge across a wide landscape of specialties, from doctors and lawyers to engineers. A self-described “deep generalist,” Amitabh certainly seems to fit the bill.
He has an undergraduate degree in mechanical engineering, a master’s degree in moral philosophy, and spent more than six years working on business development for Microsoft in a role that focused on cloud computing and AI partnerships. His past writing includes a book on “the side-hustle revolution” and a master’s thesis on how AI will affect the nature of achievement.
The opportunity with micro1 seemed like a “natural” fit, says Amitabh. He also appreciated the flexibility of the part-time, freelance role — he works, on average, roughly 3.5 hours each night, typically after his 1-year-old daughter goes to bed, he says.
“This didn’t seem like an add-on, but something that I could use to further my interests in a limited number of hours a week,” Amitabh says.
Amitabh now earns $200 per hour for his work training AI models for micro1, based on a pay stub viewed by CNBC Make It, and a company spokesman confirmed that Amitabh has earned nearly $300,000, including project completion bonuses, for his work dating back to January.
At the same time, Amitabh says that “money was less of a motivator” than the role’s overlap with his own personal and professional interests, especially considering that he already had a livable income from his other jobs, he says. Still, he considers “fair pay to be a core value,” he says, adding that he found the compensation to be “respectable” for work that requires significant expertise.
‘You need to have immense attention to detail’
Founded in 2022, micro1 has built a network of more than 2 million experts who work on training AI models for clients such as large AI labs, including Microsoft, and Fortune 100 businesses developing their own large language models for their respective workforces, according to TechCrunch. Micro1 was most recently valued at $500 million and counts larger startups like Mercor and ScaleAI among its competitors.
That network of experts like Amitabh forms the “backbone of our data quality,” micro1′s chief marketing officer, Daniel Warner, said in a statement: “Today’s AI models have already absorbed most publicly available knowledge, and real progress now comes from domain experts who can challenge, refine and effectively outthink the model. The ‘human data’ generated by true experts is what enables us to deliver best-in-class results for leading AI labs and the Fortune 100.”
AI model training involves feeding massive amounts of information and scenarios into an algorithm to form a large data set. The model is then refined over time by testing it with prompts that ask the model to answer questions or propose solutions to problems — like asking an AI agent to track expenses, project growth and create a new budget for a business unit within a company, for example.
Many of the projects he works on are confidential and involve “looking at a complex business problem that a regular user, a business owner or an executive, might have, and then breaking down that problem into small parts,” Amitabh says.
Much like prompt engineering, this part of the job requires him to break down each problem into clear, specific language that “machines will understand” to ensure the model can return an accurate and relevant response, he adds.
If there are errors in the model’s response, or it strays too far outside the parameters of the original question or problem, Amitabh works to identify where “a point got missed or subtlety got lost” and address it so the model’s data set can be adjusted and improved before testing it again. It’s a trial and error process that can take “several hours” per problem set, he says.
“You need to have immense attention to detail, and you have to often look out for mistakes that the human might make or a machine might make, and you discover more about the kinds of mistakes that exist by the process of immersing yourself in it,” Amitabh says.
The job is “intellectually quite demanding,” particularly because the AI models are constantly learning and improving, requiring even experts like Amitabh to level up their own knowledge base and creative thinking skills, he says.
“The ultimate goal is actually really energizing,” he adds. “You’re seeing whether the machine and human, the way this engagement is happening, [can] level up the output for problems that you asked and other kinds of problems that might be related to it.”
AI and jobs: ‘The trillion-dollar question’
Amid the rise of AI at work, a concern for employees across most industries is whether the advancing technology will eventually make human workers obsolete, or at least significantly transform their roles. So, does Amitabh worry that lending his own expertise to train AI models now could mean fewer career opportunities for himself, or others with similar backgrounds, in the future?
“This is the trillion-dollar question,” he says, noting that people typically fall into the camp of “techno-optimists or techno-pessimists” when it comes to how they view the impending AI revolution and its effects on the labor market. “I like to think of myself somewhere between a techno-optimist and a techno-realist,” he adds.
Amitabh concedes that there are sure to be “growing-up pains” as more companies implement AI tools in their workers’ day-to-day activities, likely resulting in the elimination of a significant number of jobs — an effect that human resources leaders say is already beginning to happen.
However, he is also in the optimistic camp that expects AI to eventually create more jobs to help offset those losses. For instance, a January 2025 analysis from the World Economic Forum predicted that AI will be a disruptive, but ultimately beneficial, force on the global labor market that will result in nearly 80 million net job gains by 2030.
Ultimately, Amitabh says he takes a more philosophical outlook: He is confident that knowledge, both in humans and machines, is not a “finite” resource, and that humans and machines will always have a symbiotic relationship where advancement for both will require perpetual collaboration.
“It’s also possible that this AI fear collectively empowers us to learn better, upskill ourselves and frame questions differently about ourselves,” he says, adding: “So I’m not concerned about the [idea of] AI Doom entirely, because I think it does far more good than bad.”
Source – https://www.cnbc.com/2025/12/19/34-year-old-entrepreneur-earns-200-an-hour-training-ai-models.html