What recruiters look for on an AI engineer CV

A recruiter reading an AI engineer CV cannot tell "used the OpenAI API" from "read about the OpenAI API." Both produce the same line on the page. The gap between them is what a list of tool names can never close, and what a description of what you built and how you checked it does close.

The rest of this piece is what that looks like in practice, so you can see whether your own CV closes the gap.

Why AI experience is a wide net

AI engineer postings vary more than most roles because AI engineering is not one job. One posting wants someone building applications on top of an LLM API. Another wants retrieval augmented generation, a system that finds relevant documents before the model answers. Another wants agents, a model taking multi step actions rather than answering once. Another wants fine tuning, adjusting a model on a specific dataset. A posting rarely wants all four.

That is usually the first thing a recruiter checks, before anything about quality: does the type of AI work in your CV match the type of AI work in this posting. A strong fine tuning background does not answer a posting asking for retrieval experience, even though both are, in the broadest sense, AI.

The tool list problem

Once the type matches, the next question is whether the tools on the CV were used or only named. A skills section listing an LLM API, an orchestration framework and a vector database is common on AI engineer CVs, and on its own it proves very little, because anyone can copy those words from a job description into a CV.

What proves something is a tool named inside a description of what it did. "Used an LLM API to sort incoming support emails into categories, so the right person saw each one first" is a tool doing real work, in a sentence a recruiter can act on. The tool name is not the evidence. The sentence around it is.

This is the same distinction the Interview Score is built around: a skill shown in context counts for more than the same skill named once in a list. How the Interview Score works sets out the method.

What built looks like on a CV

A bullet that only names what was used reads as familiarity. A bullet that says what was built reads as capability, and the difference is usually one clause. Not "experience with an orchestration framework," but what it was used to build: a support email sorter, a research summariser, a meeting notes tool. The build, named in a few words, is what turns a tool list into a project.

This does not need to be large or original. A small, clearly described personal project shows more than a vague line about a bigger one, because a recruiter can only judge what the CV actually shows.

What evaluated looks like

Beyond built, the strongest AI engineer CVs show some evidence that the candidate checked whether the thing worked. That can be as simple as one sentence: how accuracy was measured, what happened when the model produced a wrong or unsafe answer, or a tradeoff that was weighed, such as cost against response time.

This is the signal that separates a working system from a tutorial that was followed and shipped without a second look. It does not need to be rigorous by research standards. It needs to exist, in a sentence, so a recruiter can see that testing the result was part of the work rather than an afterthought.

A project counts as much as a job title

None of the above needs a job with "AI Engineer" in the title. A personal project, a hackathon build or coursework can carry exactly the same evidence, built and evaluated, and a CV that shows it gives a recruiter the same thing to weigh. What matters is the work and the checking, not the employment status behind it.

A rewrite, before and after

Here is an invented CV bullet, written for this article and not taken from any real CV.

Before: "Experience with LLM APIs and prompt engineering."

After: "Built a tool that sorts support emails into categories using an LLM API, and compared its labels with a set I had sorted by hand, rewriting the prompt for the categories it confused."

The tools have not changed. The second version says what was built, how it works in one clause, and what was checked. That is the whole difference, and it is the difference a recruiter is reading for.

A five minute self check

Take your own AI related bullets and ask three questions of each one.

• Does it say what type of AI work this is, an application, retrieval, an agent or fine tuning, or does it only say "AI"?
• Does it say what was built, or only what tool was used?
• Does it say anything about checking whether it worked?

A bullet that answers none of the three is a tool list. A bullet that answers all three is evidence.

If you want to see how your own CV reads against a specific AI engineer posting, the AI engineer CV checker compares the two and reports whether your experience is demonstrated with evidence like this or only named.