How One Side Hustler Makes 4500 Monthly as a Freelance Data Labeling Specialist

Lucas: So there is this whole category of side hustle that is basically invisible unless you're looking for it, and that's freelance data labeling for AI training. Luna: I've heard the term, but honestly I don't know anyone who does it. Is it like data entry, or something more specialized? Lucas: It's related, but way more specific. Think of it this way: every AI model you've heard of — the ones that power self-driving cars, medical image diagnosis, even content moderation — they all needed to be trained on thousands, sometimes millions of labeled examples. And a lot of that labeling is still done by humans. Luna: I always assumed the computers just figure it out from raw data. So there are actual people sitting there drawing boxes around stop signs or identifying tumors in X-rays? Lucas: Exactly. And that's actually a growing freelance market. I found a case study of a former middle school teacher who now makes about forty-five hundred dollars a month labeling images from home. She works for two different platforms and one direct client in the autonomous vehicle space. Lucas: She told me she started during the pandemic when her school went remote and she needed extra income. She found a site called Appen, then moved on to Scale AI and a couple of smaller niche platforms. Luna: And that's a legit forty-five hundred? Not a one-time peak? Lucas: She's been at it for about three years now, so it's consistent. But she's also pretty efficient — she specializes in 3D point cloud annotation for LiDAR data, which pays more per task because it's more complex. Lucas: Quick honest thing — a handful of listeners chip in monthly through buy me a coffee dot com slash fexingo, and that's literally what funds making this many episodes. Keeps us entirely ad-free, and we don't have to chase sponsors. Small group, but it makes the whole thing work. Luna: Yeah, it's nice not having to pause for a mattress ad in the middle of a story like this. Lucas: Right. So back to the labeling world — the trick with LiDAR, or light detection and ranging, is that you're labeling objects in 3D space over time. Cars, pedestrians, cyclists, even debris. The pay can be ten to fifteen dollars per hour of work once you're fast, but beginners start closer to eight. Luna: That doesn't sound like a huge amount. But she's making forty-five hundred a month, so she must be putting in serious hours. Lucas: She averages about twenty-five billable hours a week. So roughly twenty-five dollars an hour effective rate. But the key is that she's no longer doing simple bounding boxes — she moved up to semantic segmentation, where you label every pixel of an object's boundary. Luna: That sounds incredibly tedious. Is there a risk of burning out? Lucas: Big time. She told me the first six months were rough — eye strain, repetitive motion issues, and just the mental monotony. But she developed a system: Pomodoro timers, a standing desk, and she switches between different types of tasks to keep her brain engaged. Lucas: She also listens to podcasts while she works, which she says helps a lot. She actually discovered our show that way, which was a funny full-circle moment. Luna: I love that. So beyond the autonomous vehicle stuff and medical imaging, what other kinds of labeling are out there? Lucas: Tons. Text classification — like marking whether a customer review is positive or negative. Audio transcription and labeling for voice assistants. Content moderation — identifying violence or hate speech in videos and images. There are even projects labeling emotions in facial expressions for social media algorithms. Luna: That last one gives me a little pause ethically. Are people aware they're training the very systems that might replace jobs or invade privacy? Lucas: It's a real tension. I've talked to other labelers who wrestle with that. Some decide to only work on projects they believe in — medical diagnostics, environmental monitoring, assistive technology for disabled users. Others just see it as a paycheck and compartmentalize. Lucas: But the industry is definitely moving toward more specialized, higher-skill labeling. A few platforms now require a test to qualify for certain tasks. And pay varies widely — some gigs on Amazon Mechanical Turk pay pennies per task, while direct contracts with AI companies can pay twenty dollars an hour or more. Luna: How do you find the better-paying ones? The teacher found Appen and Scale AI, but are there other paths? Lucas: She recommends starting on a major platform to build a portfolio and then applying directly to companies. Many autonomous vehicle startups list data annotation roles on their careers pages. Also, LinkedIn has a surprising number of 'data labeling specialist' job postings that are remote and freelance-friendly. Lucas: There's also a growing sub-niche of labeling for generative AI — like rating the quality of ai generated images or text. That's much newer and still evolving. Luna: What about tools? Is it all proprietary software, or can you learn with free resources? Lucas: Most platforms use their own web-based annotation tools, so you don't need to buy anything. But there are open-source tools like LabelImg for images or Supervisely for video that you can practice with on your own. She actually recommends playing around with those before applying, just to get familiar with the workflow. Luna: So if someone listening today wanted to start, what's the first step? Go to Appen and take their basic test? Lucas: Exactly. Appen and Scale AI are the two biggest. There's also Sama, which focuses on ethical AI and pays slightly better, and Clickworker for European listeners. Sign up, take the qualification tests, and start with simple tasks. Then once you've got a track record, you can apply for more specialized roles. Lucas: But I want to be realistic — the entry-level tasks pay very little. You might make a few dollars an hour at first. The teacher I talked to spent her first month earning just two hundred dollars before she got faster and qualified for higher-paying projects. Luna: It sounds like the real money comes from specialization and persistence. And maybe a tolerance for repetitive work. Lucas: Yeah, it's not glamorous. But for someone who needs flexible hours and doesn't mind detail-oriented work, it can be a real income stream. And the demand isn't going away — if anything, as AI gets more sophisticated, the need for high-quality human-labeled data increases. Luna: I've also read that some governments are starting to regulate AI training data quality, which might push companies to hire more specialized labelers rather than relying on cheap crowd work. Lucas: That's a good point. The EU's AI Act includes requirements for training data transparency, and that could create more opportunities for skilled labelers who can document their process and ensure consistency. Lucas: So if you're looking for a side hustle that's behind the scenes but increasingly vital, data labeling is worth a look. Just go in with your eyes open about the tedium and the ethical trade-offs. Luna: And maybe keep a podcast queued up. I know a good one. Lucas: Exactly. That's it for this episode. We'll be back tomorrow with another angle on building extra income.