How to Write a Resume for an AI Job: What Hiring Teams Scan

An AI resume gets read when it shows real AI work with a measurable result: a model, feature or workflow you built or ran, how you checked it worked, and what changed. That holds for engineers and for sales, product or operations people. Name specific tools and methods instead of buzzwords, link proof, and put your strongest AI evidence near the top.

Definition: An AI resume is a resume aimed at roles that build, sell or run AI products, or that use AI heavily, and it's judged on evidence of hands-on AI work (shipped systems, evaluations, workflows and results) more than on listed skills.

Here's the problem. Every resume mentions AI now.

So the word itself carries little. What carries weight is specificity: which model, which data, which metric, which user. This guide covers what AI hiring teams scan for, how technical and non-technical AI resumes differ, how to write up AI projects, how to handle keywords, and how to stand out. For the broader search, our guide on how to get a job in AI covers paths in and where to look.

What AI hiring teams scan for

Demand is real, which is part of why the pile is crowded. The Stanford AI Index 2026, whose economy chapter draws on Lightcast data, reports that 2.6 percent of all US job postings required AI skills in 2025, and that postings referencing agentic AI (systems that take actions, not just answer questions) rose sharply from 2024 to 2025.

It isn't only engineering. In a July 2025 report built on more than 1.3 billion postings, Lightcast found that, as of 2024, 51 percent of postings requiring AI skills were outside IT and computer science occupations, and that jobs requiring AI skills advertised about 28 percent higher pay (nearly $18,000 a year). PwC's 2026 Global AI Jobs Barometer, which analyzed more than a billion job ads across 27 countries and territories, put the average wage premium for workers with AI skills at 62 percent, up from 57 percent a year earlier. At startups, Carta's H2 2025 compensation report found the median initial equity grant for AI/ML engineers on its platform rose 31 percent between January 2024 and February 2026, and median salary 9.1 percent. These studies measure different things, so we'd read them as direction rather than a number to quote in a salary talk.

With that much interest, readers tend to look past the label. In our view they're scanning for four things:

Did you build or run something real? Not a course. A system, a feature, a workflow someone used.
Did you measure it? Accuracy, error rate, latency, cost per request, hours saved, deals closed.
Do you understand the trade-offs? Quality vs. cost vs. speed, and where a model fails.
Can they check it? A repo, a demo, a write-up, a public eval.

Background matters less than many people assume, at least at some labs. Anthropic notes on its careers page that about half its technical staff had no prior machine learning experience, and it gives unusually direct resume advice: "If you've done interesting independent research, written a thoughtful blog post, or contributed to open source, put that at the top of your resume."

The technical AI resume: engineers, data scientists, researchers

For technical roles, the page should read like a log of systems you shipped and how you knew they worked.

What to put near the top:

Shipped AI systems, with users or traffic: a retrieval pipeline, a fine-tuned model, an agent in production, a ranking model.
Evaluation work. OpenAI's documentation describes evals as structured tests for measuring a model's performance, a way to get reliability out of systems whose outputs vary. Teams building with large language models (LLMs) tend to prize people who can design them. In Gergely Orosz's Pragmatic Engineer reporting on AI engineering (2025), one engineering leader called getting comfortable with evals and non-deterministic outputs the biggest challenge for most developers.
Cost and latency. Teams running models at scale care what each request costs and how long it takes. A bullet that moved either is strong.
Data work. Cleaning, labeling, synthetic data, pipelines. It's unglamorous and often decisive.
Research output for research roles: papers, preprints, reproductions, with your contribution named.

Illustrative before and after (the numbers are invented to show the shape):

AI engineer\ Before: Built a chatbot using LLMs and RAG.\ After: Built a retrieval-augmented support assistant over 12,000 help articles (Python, pgvector, Claude API); a 300-question eval set raised answer accuracy from 71 to 88 percent before launch, and it now resolves 34 percent of tickets without an agent.

Machine learning engineer\ Before: Worked on model training and deployment.\ After: Distilled a fraud model into a smaller version that kept 97 percent of recall while cutting inference cost per 1,000 transactions by 60 percent; deployed on Kubernetes with drift monitoring.

Data scientist moving into AI\ Before: Experimented with generative AI for marketing.\ After: Tested 3 LLM prompting approaches for product descriptions on 2,000 SKUs; the winner passed human review 92 percent of the time and cut writing time per SKU from 15 minutes to 3.

Research engineer\ Before: Researched reinforcement learning.\ After: Reproduced a published RLHF result at 1B parameters in 3 weeks, found a reward-hacking failure the paper missed, and published a write-up (link); 2 lab researchers cited the write-up.

For a deeper look at the skills behind these, our guide on how to become an AI engineer covers the stack and the projects that tend to prove it.

The non-technical AI resume: sales, product, operations, marketing

If you're applying to an AI company in a go-to-market or operations role, or to a role that expects heavy AI use, the page has two jobs. Prove you're strong at the function. Then prove you can work with AI, not just talk about it.

The second part is where most resumes go vague. "Proficient in ChatGPT" tells a reader very little. A workflow with a result tells them a lot.

Account executive at an AI company\ Before: Sold AI solutions to enterprise customers.\ After: Closed 6 enterprise pilots of an AI document-review tool; built ROI cases from each customer's own review hours, and 4 pilots converted to annual contracts worth $1.1M.

Operations manager\ Before: Used AI tools to improve efficiency.\ After: Automated vendor invoice matching with an LLM workflow and a human check on exceptions; cut monthly close work by 40 hours and caught 23 duplicate payments in the first quarter.

Product manager\ Before: Led AI feature development.\ After: Shipped an AI meeting-summary feature to 8,000 users; defined the quality rubric and weekly eval review with engineering, and kept thumbs-down feedback under 6 percent after launch.

Marketer\ Before: Leveraged generative AI for content.\ After: Built a prompt library and review checklist that let a 3-person team publish 3x more comparison pages; organic signups from those pages rose 45 percent in two quarters.

Notice the pattern: a real task, AI used in a specific way, a human check where it matters, a number. If you're mapping which non-coding AI roles fit your background, our guide to non-technical AI jobs covers the options.

How to list AI projects on a resume

Projects carry extra weight in AI because the field moves fast and many people's best work isn't in their job title yet. A project entry that works, in our view, has five parts. Copy and adapt:

Project name | one-line description | link

Problem: who had it and why it mattered
Build: model or API, data, tools (be specific)
Evaluation: how you tested it and the score
Result: users, time saved, cost, accuracy, or what you learned
Constraint: the trade-off you made (cost, latency, privacy)

A filled example (illustrative):

Lease Checker | flags risky clauses in residential leases | GitHub and live demo

Built for a tenants' rights volunteer group handling about 50 lease reviews a month
Claude API with structured outputs, a clause taxonomy from 40 annotated leases, Next.js front end
120-clause test set: 91 percent recall on high-risk clauses, checked against a volunteer attorney's labels
Volunteers now pre-screen leases in about 10 minutes instead of 45
Kept documents out of logs and cut cost per lease to under 5 cents with a smaller model for triage

That last line is the one many candidates skip. It's often the one that tells a reader you've done this for real.

Two or three strong projects tend to beat eight thin ones. Course assignments and tutorial clones tend to read as homework unless you pushed them somewhere new.

AI resume keywords, without stuffing

Keywords still matter, because recruiters search applicant tracking systems for terms. Our guide on how to write a resume for a tech job covers how that software actually works. But in AI, the vague keyword is often the weak one.

Vague	More specific (use only if true)
AI, machine learning	Classification, ranking, forecasting, the model type
LLMs	The model families and APIs you've used
Generative AI	RAG, fine-tuning, structured outputs, agents
Prompt engineering	Eval-driven prompt iteration, with the metric
AI tools	The named tools, plus what you did with them

A few rules of thumb:

Mirror the posting's terms where they're true. If it uses "evals" and you built test sets, use evals.
Put each key term in a bullet with a result, not just a skills list.
Skip terms you can't discuss for ten minutes. AI interviews tend to probe; our guide on how to prepare for an AI job interview shows how.
Date-check your stack. Model names change fast; a list frozen two years ago can read as stale.

Should you use AI to write an AI resume?

It's a fair question when you're applying to the companies that make these tools. Some employers spell out their expectations. Anthropic's candidate guidance asks applicants to write the first draft themselves and then use Claude to refine it, and asks that live interviews and take-home assessments be done without AI unless the company indicates otherwise.

That seems like a sensible default anywhere. Use AI to tighten wording, check for gaps against a posting and catch errors. Keep the substance yours. Readers in AI tend to spot generated filler quickly, and claims you can't defend in an interview tend to unravel there.

"AI is on every resume now, so how do I stand out?"

Fair. When everyone lists the same tools, the list stops working as a signal.

But the way to stand out for AI jobs isn't more AI words. It's narrower, deeper proof:

One public, evaluated project in the domain of the company you want. A legal AI startup will read a lease-review project very differently from a generic chatbot.
A short write-up of what failed and how you fixed it. It shows judgment, which is hard to fake.
Domain expertise plus AI. A nurse who built a triage-note eval, or an accountant who automated reconciliations, often brings something a pure engineer can't.
Timing. Apply while roles are fresh. The AI track on the 1752vc careers board lists open roles at AI companies with each employer's posting date, and you can filter by level and time posted.

Our view: a resume with two specific, measured AI results usually beats one with twenty AI keywords.

Common AI resume mistakes

Buzzword bullets. "Leveraged cutting-edge AI to drive innovation" gives a reader nothing to check.
Projects without evaluation. "Built a RAG app" invites the question: how good was it?
Inflated claims. "Built an LLM" when you called an API is easy to catch in a first interview.
No links, or links to empty repos.
Ignoring the function. For non-technical roles, AI skills sit on top of sales, product or ops results. They don't replace them.
One resume for every AI role. A research lab, an applied AI startup and a company adopting AI read the same page differently.

Where we land

An AI resume is a short record of AI work you can prove, not a list of AI words you've heard. Technical or not, show what you built or ran, how you measured it, what it cost and what changed. Then link it.

That's our approach, and AI hiring is moving quickly, so expect norms to shift. When you're ready to apply, the AI jobs track is a place to find roles at AI companies that went up recently.

The bottom line

Anyone can type "AI" into a skills section.

Fewer people can show you the eval.

Key takeaways

Strong AI resumes show hands-on AI work with a measurable result (accuracy, cost, latency, time saved or revenue), not lists of AI terms.
AI skills are spreading well beyond engineering; Lightcast found that, as of 2024, 51 percent of postings requiring AI skills were outside IT and computer science occupations.
Technical AI resumes tend to lead with shipped systems, evaluation work, cost and latency trade-offs, and data work.
Non-technical AI resumes work best when AI sits on top of strong function results, shown through specific workflows with human checks.
Specific terms in result-backed bullets beat broad keywords, and AI writing help works best on a draft of your own.

Frequently asked questions

What should an AI engineer put on a resume?

Lead with AI systems you shipped and how you knew they worked: users or traffic, the evaluation set and score, and any cost or latency gains. Name the models, APIs, data stores and frameworks you used. Add two or three linked projects if your job doesn't show AI work yet, and keep the skills list to tools you can discuss.

How do I list AI projects on my resume?

Give each project a name, a one-line description and a link, then cover the problem, what you built and with which tools, how you evaluated it, the result and one trade-off you made. Two or three well-documented projects usually read better than many thin ones, especially if one matches the domain of the company you're applying to.

What keywords should an AI resume include?

Use the specific terms from the posting that truthfully describe your work, such as RAG, fine-tuning, evals, agents, or named models and frameworks. Place them inside bullets that show a result, not only in a skills list. Vague words like "AI" or "generative AI" on their own tend to carry little weight with technical readers.

Is it okay to use ChatGPT or Claude to write a resume for an AI job?

Some employers openly accept AI help with polishing. Anthropic's candidate guidance asks applicants to write their own first draft and then use Claude to refine it. Keep the facts and substance yours, check every claim, and expect interviewers to probe anything on the page, especially at AI companies.

Do you need machine learning experience to get hired at an AI lab?

Not always. Per Anthropic's careers page, about half its technical staff had no prior machine learning experience, and the company cares about what you can do rather than where you learned it. Strong engineering, research, open-source or writing evidence can carry a resume, though specialized research roles often still favor deep ML backgrounds.

Sources

Disclaimer: This guide is for general education only and is not legal, tax or investment advice. Laws, market data and program terms change, so it may not reflect the latest developments or fit your situation. Treat it as a starting point, not a source of truth, and talk to a qualified lawyer, accountant or financial adviser before you make decisions.