Latest / The Side Hustle Career Podcast with Fexingo: Second Income, Freelancing While Employed / How a Retail Manager Built a 5000 Monthly Side Hustle in Data Annotation
Transcript
- Lucas: So you've heard of AI models like ChatGPT or Claude. But here's a question — how do they actually learn to recognize a stop sign in an image or understand that 'bank' can mean a riverbank or a financial institution? Luna: I mean, broadly speaking, training data. But I think for most people, that's just a black box. Lucas: It is. And inside that black box is a huge amount of human labor. Specifically, data annotation — the work of labeling images, transcribing audio, or ranking search results so the AI can learn patterns. And it's become a legit side hustle for a lot of people. Luna: So for today's episode, you've got a specific story? Lucas: Yeah. I talked to a woman named Jen. She's a retail manager in Ohio, makes about forty-five thousand a year at her day job. Last year, she started doing data annotation on the side — mostly image labeling and text categorization — and within six months, she was averaging five thousand dollars a month. Working about twelve hours a week. Luna: Twelve hours a week? That's a pretty solid hourly rate. Lucas: It works out to around a hundred dollars an hour. But here's the catch — it's piecework. You get paid per task, not per hour. And the tasks vary wildly in complexity. Jen's sweet spot was what the platforms call 'image segmentation' — tracing objects in photos with a pixel-level mask. It pays well because it requires careful attention. Luna: That sounds tedious. But I guess if you're playing a podcast or watching TV while you do it, it's tolerable. Lucas: Exactly. Jen told me she treats it like a puzzle. She puts on a show in the background and just zones in. And the barrier to entry is really low. You don't need a degree in computer science. You need basic English literacy, attention to detail, and a willingness to follow guidelines. Luna: So what kind of tasks are we talking about? Give me a concrete example. Lucas: One common task is bounding boxes. The platform shows you an image — say, a street scene — and asks you to draw a box around every car, pedestrian, and traffic light. Another is sentiment analysis: read a customer review and classify it as positive, negative, or neutral. Or you might transcribe a short audio clip of someone saying 'turn left at the gas station'. Luna: So it's basically teaching the AI what things are. Lucas: Exactly. And the demand has exploded. Companies like Scale AI, Appen, Labelbox, and Amazon's Mechanical Turk — they all need a steady stream of human annotators. Scale AI alone reported revenue of over a billion dollars last year, mostly from providing training data for autonomous vehicles and large language models. Luna: Okay, but a lot of these platforms have a reputation for paying pennies. I've heard horror stories about Mechanical Turk paying two dollars an hour. Lucas: That's real. The lower-end tasks on Mechanical Turk can pay terribly. But Jen specifically avoided those. She qualified for more specialized projects — things like Lidar annotation for self-driving cars or medical image labeling. Those pay significantly more because they require more judgment and the client is willing to pay for accuracy. Luna: How did she qualify for those? Sounds like there's a vetting process. Lucas: Most platforms have a qualification test. You take a short training module, then do a handful of sample tasks. If you pass a certain accuracy threshold — say, ninety-five percent — you get access to the higher-paying projects. Jen spent about two weeks grinding through qualifications before she hit the good stuff. Luna: So there's an upfront investment of unpaid time. That's important for listeners to know. Lucas: Absolutely. It's not instant money. But once you're in, the work can be steady. Jen said she never experienced a dry week after the first month. She checks for new tasks every morning and picks up batches that fit her schedule. Luna: And she's doing this while managing a retail store? That's impressive. How does she find the energy? Lucas: She does it in short bursts. Fifteen minutes here, thirty minutes there. She keeps the annotation tool open on her personal laptop and tackles tasks during her lunch break or after dinner. She told me she treats it like a habit, not a grind. Luna: I like that framing. So for someone listening who wants to try this, what's the first step? Lucas: I'd say start with the bigger platforms that have a track record. Scale AI has a program called 'Scale Data Engine' that hires independent contractors. Appen and Lionbridge are two others that have been around for years. Also, check out 'Clickworker' — it's European but has global projects. Avoid any platform that asks you to pay a registration fee. That's always a scam. Luna: Good rule of thumb. Any specific skills that make someone more competitive for the higher-paying tasks? Lucas: Yes. If you have experience in a niche domain — like medical terminology, legal language, or an Asian or Middle Eastern language — you can command premium rates. Jen doesn't have any of that, but she's very consistent. Her accuracy scores are in the ninety-eighth percentile, so project managers give her priority access to new tasks. Luna: So it's about reliability. That makes sense. The AI companies need data they can trust. Lucas: Exactly. And here's the thing — this work is often seen as a stepping stone to a career in AI. But Jen has no interest in that. She just wants the extra income to pay down debt and save for a down payment on a house. And she's on track to earn about sixty thousand from annotation this year. Combined with her retail salary, that's over a hundred thousand. Luna: That's a life-changing amount for a side hustle. And it shows you don't have to be a tech bro in San Francisco to benefit from the AI boom. Lucas: Not at all. And it's not just about money — it's also about flexibility. Jen can work from her couch, in her pajamas, at three in the morning if she wants. There's no boss hovering, no meetings. Just the tasks and her own discipline. Luna: Alright, so we've painted a pretty rosy picture. But what are the downsides? I want to be realistic here. Lucas: The biggest downside is that the work can be mind-numbing. Doing bounding boxes for two hours straight can give you a headache. And your eyes get tired. Jen takes a five-minute break every twenty-five minutes — she uses a timer. Also, the income is not guaranteed. If a big client pulls their project, tasks can dry up. She had one slow week in February when a major contract ended. Luna: So it's not passive income. You have to actively work for every dollar. Lucas: Correct. It's a gig, not an investment. But for someone who has a few hours a week and wants predictable extra cash, it's one of the most accessible options right now. Luna: And speaking of accessible — one thing I love about this show is that we don't clutter it with ads. You know, we're not trying to sell you a course or a subscription. That's intentional. Lucas: Yeah, we've always believed that career advice should be free from sponsored pitches. The only reason we can keep it that way is because some listeners choose to support us directly. If you've ever gotten value from an episode like this, the link is buy me a coffee dot com slash fexingo. It's a simple way to chip in and help us stay ad-free. Luna: And we genuinely appreciate it. It means we can keep focusing on what matters — real stories and practical steps. Lucas: So, back to data annotation. One thing I want to emphasize is that this isn't a get rich quick scheme. It took Jen months to build up to five thousand a month. But if you're willing to put in the upfront time to qualify and build a reputation, the upside is real. Luna: And for anyone worried about AI taking over annotation jobs — ironically, human annotators are still essential. The AI needs to learn from humans first. There's a phrase in the industry: 'human-in-the-loop.' For now, at least, the loop still needs us. Lucas: That's a great point. Until AI can reliably judge its own output — which is a long way off — there will be a need for human annotators. So this isn't a fad. It's a growing field. Jen told me she sees it as a part-time job that could last for years. Luna: I want to ask about the tax implications. She's a contractor, right? So she has to handle self-employment tax. Lucas: Yes. She sets aside about thirty percent of every payment in a separate savings account. She also tracks her internet bill and a portion of her rent as a home-office deduction. She uses a simple spreadsheet. If you're not comfortable with that, she recommends hiring a freelance accountant for a one-time session — cost her about two hundred dollars. Luna: Good advice. So for our listeners who want to try it, what's the one concrete action they can take today? Lucas: Go to Scale AI's website and look for their 'Data Annotation' roles. Sign up, take the qualification tests, and start with the lower-paying tasks to build your accuracy. From there, you can work your way up. And don't quit your day job yet — treat it as a trial for a month. If after thirty days you've made less than a few hundred dollars, maybe it's not for you. But if you stick with it, like Jen did, it could become a serious income stream. Luna: That's fair. I think a lot of people are looking for a side hustle that doesn't require selling or creating content. This is pure task-based work. It's almost like a modern version of piecework from the industrial era. Lucas: Exactly. And it's worth a shot if you have the patience. Jen said the first week she made just forty-two dollars. But she kept going, and now she's on track to make more from her side hustle than from her full-time job. Luna: That's a powerful testament to persistence. Alright, so next episode, I'd love to explore another angle — maybe how people are using AI tools themselves to automate parts of their side hustle. Lucas: That's a great idea. We'll get on that. For now, thanks for listening, and we'll catch you next time.