Can AI really help land a job?

If you’re reading this, you’re probably a bit like me: endlessly curious, occasionally stubborn, and always up for a bit of tinkering—whether that’s with code, data, or the odd hiking boot. By day, I wrangle data & analytics for a living; by night (and the odd lunch break), I dive into the world of AI agents and coding tools, equal parts professional and hobbyist, always learning, sometimes swearing at my screen.

You might also be like me and find yourself searching for the next work adventure. If you’re feeling frustrated by the faceless gatekeepers blocking your way to the interview stage, you’re not alone.

The Applicant Tracking System (ATS) may seem like a godsend for employers, but it’s also quietly discarding many excellent resumes from people who could genuinely make a difference. Ever spent hours tweaking your CV, only to get an automated rejection in minutes? Welcome to the club.

Sure, some recruitment firms offer to “optimise” your resume for the ATS, but what happens when we all have the same keywords and bullet points? If every resume looks the same, does anyone really stand out?

So, what can you actually do to get noticed? People can spend hours tailoring resumes for each application, cross-referencing job ads for that elusive keyword—only for the bots to filter them out without a second glance. It’s enough to make even the most optimistic job-seeker feel invisible.

This article is about taking some control back: using AI to identify your resume’s strengths and weaknesses, compared to the real requirements of the jobs you’re applying for. I’ve been on a mission to reverse-engineer the ATS profile for any position—asking AI to analyze real job descriptions, build an ideal candidate profile, and then compare that to your resume, providing AI-powered guidance.

And, because I value privacy as much as a well-brewed coffee, the solution I built lets you self-host the app. Your search history and resumes stay in your browser—no recruitment agency harvesting your details for their own database.

So, allow me to introduce one of my proudest side projects: the ATS Resume Optimizer. I’ll walk you through how it was built, why it matters, and how it could help you outsmart those enigmatic, sometimes maddening, Applicant Tracking Systems. (Spoiler: you’re more than just a keyword to a robot, and I’ll show you why.)

Context & Motivation

My journey started with about 90 seconds of a YouTube video showing the Copilot agent in Visual Studio Code. I watched in awe as it spun up a new app and fixed its own errors from just a few simple instructions. Cue my curiosity.

I’ve dabbled with simpler projects before, but this time I wanted to see how far a non-professional coder—armed with curiosity and caffeine—could push things. Could I build a prototype that’s genuinely useful, not just a digital doodle? It turns out, yes. And it’s a game-changer for brainstorming, prototyping, and sharing real solutions with project stakeholders (instead of the usual napkin sketches).

The Learning Curve

We’ve all tested AI models to see what they’re really capable of—whether it’s generating an image, rephrasing a paragraph, or helping with a gnarly piece of code. But I wanted to dig deeper: which tools would actually help a non-professional developer like me?

Visual Studio Code, Copilot Extension & Claude 4 Sonnet

This was my launchpad, inspired by that YouTube clip. Setup was a breeze. I fed in my app description, and watched as it scaffolded a React app before my eyes.

My first experiment? An interactive calendar that, when you pick a date, shows facts for that day—holidays, UK trivia, celebrity birthdays. The prompt was basic at first, but the AI handled my tweaks and suggestions like a pro. Watching it edit code in real time was like having a tireless coding buddy who doesn’t mind my constant interruptions.

I even asked it to create a docker-compose file with a Tailscale sidecar—just to see if it could. (Spoiler: it could.)

Did I provide any data sources? Nope. I wanted to see how much it could figure out on its own. The end result was surprisingly solid—something that would have taken me much longer solo.

Next, I tried a calendar app with event history stored in PostgreSQL. Again, a few hints about the data model, and off it went. Guiding the AI felt like coaching a very eager intern: sometimes you have to remind it to focus, but it gets the job done.

Warp & Claude 4 Sonnet

Switching tools, I tried to build a resume analyser with Warp and Claude 4 Sonnet. The idea was to cross-reference a resume against job descriptions.

Did it work? Not quite. Warp was great at writing code and summarising changes, but it kept getting sidetracked—like a puppy chasing too many tennis balls. Mock data crept into the results (why would an HR Director need git?), and it started categorising keywords in ways that made less and less sense.

When I tried to scale up (profiling 50 jobs), it wanted to go down a web scraping rabbit hole. At that point, I realized I needed to be more direct with my prompts—baby steps, not leaps.

Google AI Studio & Gemini 2.5

This combo was a breath of fresh air. I revisited the resume analysis app, using the same basic description as before.

What stood out? The development environment: assistant, code, and preview panes, all side by side. The AI suggested using itself to research job descriptions, rather than scraping the web, and even structured the prompts for Gemini to follow.

Here’s a taste of the prompt (paraphrased for brevity):

“You’re an expert recruitment analyst with access to Google Search. Analyze the current job market for ‘${jobTitle}’, synthesize the top skills, qualifications, and keywords, and return a JSON object sorted by frequency.”

The AI not only defined what to research, but also how to format the results. No mock data, no guesswork—just live research, cached in the browser for easy comparisons.

Firebase Studio & Gemini 2.5

With this duo, I built a web app to display sports scores for a local club—rotating through today’s games, past matches, league tables, and fixtures.

Sometimes, I had to provide more detail on the API definition, but the AI was quick to adapt. One quirk: it liked to “fix” everything at once, sometimes breaking more than it repaired. Lesson learned—baby steps are your friend.

Spotlight: ATS Resume Optimizer

This project was a joy. The AI helped me think through the design, acting more like a collaborator than a critic.

The app lets you ask AI to research a position you’re interested in. You can use cached searches or update them. It’s not the fastest tool in the shed, but it’s live and thorough.

No surprise, my resume isn’t a great fit for HR Manager—but the AI’s research made that clear. Here’s the analysis for a Data and Analytics Architect:

By reverse-engineering the ATS profile from multiple live job ads, you get a better shot at adapting your resume—not just for one job, but for a whole category of opportunities.

Why This Matters

Job hunting is tough, and often a lonely slog. You can spend hours applying, only to hear nothing back. It’s easy to feel invisible, no matter how hard you try.

My goal is to help you feel a little more in control. You’re not handing your resume over to a bot for a generic rewrite—you’re using real insights and still choosing what to keep or change. Your resume should still sound like you. That’s what matters.

Maybe you’ll still go through the ATS machine, but with a few new tricks up your sleeve, you might just stand out.

Takeaways & Tips

• Get hands-on: Building and testing in real time is the fastest way to learn what works.

• Be specific: Macro-level requests are good, but detail about APIs and data models saves time—even if you’ve already linked documentation.

• Be kind to AI: You never know when it’ll remember you. (And it’s always polite to your future robot overlords.)

• Checkpoint your code: Export or download regularly. I once had to roll back 45 minutes of “helpful” AI fixes that broke more than they solved.

• AI loves to flatter: Take compliments with a grain of salt and always question the suggestions—AI rarely challenges your logic.

• Share your knowledge: The code for this project is available here:
https://github.com/1timnightingale/ATS-Resume-Optimizer ↗

• Yes, I know the spelling is American!

Disclaimers

• I’m not a recruiter, so this is just one perspective.

• No one paid me to write this article.

• AI isn’t always right (as I found out), so check any code or suggestions before using them.

• This project is for education. Always use your best judgment when tailoring your resume—AI insights are handy, but not a substitute for professional advice.

The Trek Continues

I’m always learning, and sharing my experiences is something I value. This was my first real attempt at building with AI—figuring out what works, what doesn’t, and what to watch out for. Years of deploying data and analytics solutions taught me that new tech always comes with quirks, especially around security and organisational policies.

As I keep exploring, I’ll share what I learn. If you try the app or have your own AI adventures, I’d love to hear from you. Let’s keep tinkering and trekking—together.