From rejection to innovation: How I built a tool to beat AI hiring algorithms at their own game

Job hunting in 2024 feels like trying to solve a TV murder mystery without watching any of the episodes, or playing a game of darts with a blindfold on. 

After months of never-ending applications, dead-end emails and deafening silences, I was left with one burning question: 

Isn’t there a better way?

As it turns out, there may be. And sometimes, frustration is the perfect spark for innovation. 

This is the story of how my own struggles with job hunting led me to build an AI-powered resume analyzer — a tool designed to help job seekers navigate the increasingly automated hiring process. But it’s also a story about resilience, creativity and the power of turning frustration into action.

Investigating the systems that rejected me

For months, I found myself caught in the unyielding cycle of sending out resumes and receiving no responses. I tailored, tweaked and optimized my applications, following every piece of advice I could find. Yet, rejection — or worse, no response at all — was the norm. Like so many others, I felt invisible, as if my hard work and potential were stuck behind an impassable wall of automated filters.

Eventually, I started investigating the systems rejecting me. 

Applicant tracking systems (ATS) emerged as the primary culprit — keyword-driven algorithms deciding whether I was worth a second glance. Even perfectly qualified candidates like me were being filtered out because our resumes didn’t “speak the language” of these systems. 

This is far from just a tech sector issue. None of the companies in a recent survey of high-volume industries said they avoid AI in hiring entirely, according to HR tech firm Phenom — we’re talking about retail, hospitality and travel, healthcare, manufacturing and transportation and distribution.

Frustrated but determined, I began tinkering with solutions. I wrote scripts to analyze job descriptions and compare them with resumes, looking for patterns. I manually optimized my resume with this data and started seeing results. A trickle of interviews, followed by more. 

The more I learned about ATS and the challenges job seekers faced, the clearer it became: I needed to build something better.

Working toward a lean, mean, resume-analysis machine

Turning an idea into a functional app was both daunting and exciting. My goal wasn’t just to create another resume tool but to design something empathetic, intuitive and genuinely useful. 

Early experiments with keyword-matching algorithms led to modest successes, but they were limited. So, in what now seems like an obvious step, I decided to see if I could improve my responses by integrating large language models like those from OpenAI, specifically GPT-3.5-turbo and GPT-4-turbo.  I used these models to provide tailored feedback, identify ATS-friendly keywords and analyze resumes against job descriptions in seconds.

The journey wasn’t without hurdles. Building features that worked across diverse industries while keeping the interface simple required constant iteration. 

At every step, I was driven by one question: How do I make this tool as helpful as possible? That meant balancing functionality with affordability, keeping the tool lean enough for job seekers yet robust enough to deliver useful responses. Along the way, I added caching to reduce costs, refined keyword-matching algorithms for accuracy and incorporated feedback from beta testers (frequently my wife Yasmin, who deserves a shoutout for her endless patience and support!).

The hardest part was embracing imperfection. As the kids say, just ship it!  

Early versions weren’t flawless, but they worked. Each setback became an opportunity to improve. As more people tested the tool, their stories reminded me why I started: to empower job seekers to overcome an unfair system.

Community (and feedback) will continue to be key

Building this tool taught me lessons far beyond coding. First, resilience matters — in building a product and in facing life’s challenges. Second, community is everything. The encouragement, feedback and support I received made the impossible feel achievable. Finally, I learned that frustration can be a powerful motivation for personal growth, and change, if we let it.

Today the tool is live, and I’m excited about its potential to help job seekers around the world. While it’s still in its early stages, I’m hopeful it can become a valuable resource for those navigating the challenges of the modern job market.

And this is just the beginning. I’m actively working to expand the tool’s capabilities, improve the analysis and keyword matching, enable access to other AI models like Claude, and create a Chrome extension to make the user experience even more streamlined.

This tool was built for people like us — job seekers who deserve to be seen, heard and given a chance. And if my experience has taught me anything, it’s this: with creativity, persistence and courage, you can turn frustration into opportunity, and have fun along the way!

Want to try the tool or share your feedback? You can explore it here  (the app uses Google sign-in, and offers 10 free scans a month). Together, let’s make job hunting a little less overwhelming and a lot more hopeful.