Most advice about “getting into AI” is really about deciding to learn the field or picking a curriculum. This is about the part that comes after that: turning self-taught skill into an actual job offer. That’s a different problem. You can know the material and still get filtered out at the resume stage, freeze up when an interviewer asks about your “professional experience,” or waste months applying to roles that were never realistic first steps. This guide is about the job hunt itself.

What to put on a resume when you have no formal AI experience

A resume built around courses you’ve taken reads like a to-do list, not a set of qualifications. Recruiters and the applicant tracking systems that screen resumes before a human ever sees them are both looking for evidence you can do the work, not evidence you’ve watched someone explain it. That changes what belongs at the top of the page.

Lead with a skills section organized around what you can actually do: which languages and tools you’re comfortable with, which kind of problems you’ve solved (classification, text analysis, basic model evaluation), and any relevant software you use daily. Follow it with a short projects section, two or three entries, each with one line on the problem, one line on what you built, and a link to the code. If you’ve held any job at all, even unrelated, keep it on the resume and describe the transferable pieces: working with data, meeting deadlines, communicating with non-technical people. Hiring managers for entry-level roles read that as evidence you can hold down a job, which matters more than people assume.

Skip the objective statement at the top. It rarely says anything a hiring manager couldn’t guess from the fact that you applied.

Build a portfolio that does the talking for you

A certificate tells an employer you finished something. A portfolio tells them you can do something, which is the actual question they’re trying to answer. The gap between those two is where most self-taught applicants lose ground without realizing it.

Two or three finished, well-documented projects beat ten half-finished ones. Pick projects tied to a real, specific problem rather than a straight rebuild of a tutorial: something like classifying support tickets by topic, predicting a simple outcome from a public dataset, or building a small tool that summarizes or tags text. For each one, write a short README that states the problem in one sentence, what you tried, what worked, and what you’d do differently with more time. That last part matters more than people expect. It shows judgment, not just output.

Put the code on GitHub with clear commit history, not a single upload dumped in one folder. Kaggle is a reasonable place to practice on real datasets and see how other people approached the same problem, but a Kaggle notebook alone isn’t a portfolio. Pull the strongest one or two into your own repository, clean them up, and write them up in your own words. If you want a simple home base to point people to, a one-page portfolio site with your projects, a short bio, and a way to contact you is enough. It does not need to be elaborate.

Which free certificates are worth listing, and which just add clutter

Free certificates vary a lot in how much weight they carry, and listing the wrong ones can make a resume look padded rather than credible. A short course completion badge from a generic platform tells an employer almost nothing on its own. A certificate tied to a recognized provider, paired with an actual project you can point to, tells them more.

• Worth listing: certificates from well-known providers when paired with a project that proves the skill, and anything that involved a real graded assessment rather than just watching videos to completion.
• Worth skipping: a long list of short course badges with no project behind them. Five one-hour certificates read as clutter, not depth.
• Worth mentioning briefly: structured programs that took real, sustained effort, even without a formal certificate, if you can describe specifically what you built during them.

For a fuller breakdown of which free certificates actually carry weight, see our guide to free AI courses with certificates.

Talking about self-taught skills in an interview without sounding unsure

The instinct when you’re self-taught is to either downplay it or over-explain it. Both read as uncertain. The better approach is to state it plainly, once, and then move straight into evidence.

Say directly that your background is self-taught, and follow it immediately with a specific project: what problem it solved, one decision you made along the way, and one thing that didn’t work the first time and how you fixed it. That last part is the piece people leave out, and it’s usually the most convincing. Anyone can describe a project that went smoothly. Describing what broke and how you debugged it is what separates someone who followed a tutorial from someone who actually understands what they built.

When you don’t know something in an interview, say so plainly and describe how you’d go about learning it rather than guessing. Interviewers screening self-taught candidates are often testing for exactly this: whether you can be honest about the edges of your knowledge without freezing up. Being able to explain your process, not just your output, is usually worth more than trying to sound like you already know everything.

Realistic entry points, not the job title you’ll eventually want

Aiming straight for “AI research scientist” or “machine learning engineer” with no formal background and no portfolio yet is the single most common way people burn months on applications that were never realistic first steps. Those roles usually expect a track record you don’t have yet. There’s a tier of roles below them that hire self-taught and career-entry candidates directly, and they’re a genuinely useful way in, not a consolation prize.

• AI data annotator or labeler: tagging and labeling text, images, or audio so a model can learn from it. Little to no coding required, and it’s a real way to see how training data actually shapes a model.
• AI trainer or model evaluator: reviewing model outputs, rating quality, and writing or ranking prompts, often for companies building or fine-tuning language models. Domain knowledge from a previous career can matter more here than a coding background.
• Prompt tester or QA for AI products: testing how an AI feature behaves, documenting failures, and describing them clearly enough for an engineer to fix.
• Junior or entry-level data analyst: working with data day to day, which builds exactly the foundation core machine learning roles later require.
• Support, operations, or research-adjacent roles that list AI tools as a requirement: these use your existing field knowledge as the differentiator, with AI fluency as an add-on rather than the whole job.

Each of these is a genuine foothold, not a dead end. People move from data annotation into more technical roles once they have real, on-the-job exposure to how models are trained and evaluated. Treat the first role as a place to build a track record, not as the ceiling.

Where to actually look for these roles

General job boards work, but searching “AI” on them mostly surfaces senior roles that assume years of experience. A more useful approach is to search by task rather than by title: “data annotation,” “model evaluation,” “AI trainer,” or “prompt tester” turn up openings that plain “AI job” searches miss entirely.

LinkedIn is worth using for more than applying. Following people who work in roles you’re targeting, commenting thoughtfully on posts about projects similar to yours, and posting your own project write-ups puts you in front of people who hire, not just algorithms that screen you out. Indeed and other general boards remain useful for volume searching once you know which titles to filter for. GitHub itself functions as a discovery channel too. Recruiters at smaller companies and startups do search GitHub directly for people with relevant, visible projects, which is one more reason a clean, documented repository is worth the time it takes to build.

If you’re still deciding whether a full career switch into AI makes sense before you commit to this job search, our guide on free AI courses for career changers covers that earlier decision. And if you haven’t settled on which skills to build first, our AI skills and career roadmap lays out the order that tends to work.

A realistic timeline for the search itself

Building two or three solid projects and a clean resume typically takes several weeks of focused effort on top of whatever learning you’re already doing. After that, expect the application and interview process to take longer than the learning did, often a couple of months of steady applying, adjusting your materials based on what gets responses, and following up. That’s normal for any competitive entry-level search, not a sign that your approach isn’t working. The applicants who land something are usually the ones who kept refining their pitch and their projects while they searched, rather than sending the same resume out unchanged for months.

Frequently asked questions