Why I Built an AI Recruiter That Shows Its Work

The story of HCIA, and a one-person bet that hiring can be both faster and more human.

The Spark

I've been writing code for over a decade. I was good at it. So I didn't expect job hunting to turn into the hardest problem I'd ever try to solve.

Every application felt the same. Polish the resume, hit submit, watch it disappear. No reply, no reason, just silence. And I had a pretty good idea what was going on behind that wall, because I'd seen it from the inside. Recruiters buried under hundreds of applicants, leaning on gut feeling or keyword tools that rewarded buzzwords over actual skill. Good people got filtered out for writing "developed" instead of "engineered." Weaker ones made it through because they'd memorized the right words.

That whole experience became the seed for HCIA.

The Challenge

Building an AI recruitment platform on my own meant doing every job at once. Product manager in the morning, backend engineer in the afternoon, designer and marketer at night. But the workload was never really the hard part. The hard part was a question I couldn't shake.

The market was already full of AI hiring tools, and most of them had the same problem. They hid their reasoning. A score showed up, a candidate got ranked, and nobody could tell you why. Not the recruiter, not the person who applied. When a tool is helping decide who gets a shot at a job, "just trust it" doesn't cut it. I didn't want to build something that made bad decisions faster. I wanted hiring to be more open than it had ever been, even back before AI was involved at all.

Building the Core

I started with the part that hurt the most, which was CV screening.

The point wasn't only to pull data out of a resume. It was to explain a match in a way a person could check. HCIA reads a CV and picks out the structured stuff: skills, experience, education, projects. Then it lines each one up against what the role actually asks for. And every match comes with a reason written in plain language:

"Candidate has 5 years of Python experience, matching your requirement of 3+ years."

No mystery number. Just the evidence.

After that came AI-assisted video interviews. Candidates record answers to a set of preset questions whenever it suits them. HCIA looks at each answer across three things, content, delivery, and sentiment, and again, it explains itself:

"The candidate addressed all three key points with confidence (77% confidence score)."

Everything it says points back to something real. No black boxes.

Under the Hood

Underneath it all, HCIA uses natural language processing to read resumes and transcribe interviews, transformer models to understand what someone actually means instead of just matching keywords, and a scoring layer that's built to be looked at rather than taken on faith.

The toughest call I made wasn't picking a model. It was deciding not to ship anything the system couldn't explain. If a feature made the scores a bit more accurate but turned the reasoning into a black box, it didn't make it in. That one rule ended up shaping pretty much every technical decision that came after it.

Explainability Is the Foundation

A lot of products treat explainability as something you add on later. For HCIA it's the thing everything else sits on top of.

Every decision the system makes points back to specific evidence, a line in a CV or a moment in an interview. Recruiters can open up any score, see what drove it, override it, or ask for more. The AI doesn't get the final word. The person does. That matters for a few reasons. It builds trust with the people doing the hiring. It gives candidates a process that's actually fair and that they could push back on. And it keeps companies on the right side of rules like the EU AI Act and GDPR, which are starting to require that automated decisions can be explained.

In hiring, "the algorithm decided" is turning into a real legal risk. HCIA is built so nobody ever has to say it.

Facing Bias Head-On

You can't talk about AI in hiring without talking about bias. Models learn from past data, and past hiring data is full of human bias. An opaque system doesn't get rid of that bias. It just hides it behind a number that looks neutral.

This is another place where explainability earns its keep. Because every decision can be traced, bias becomes something you can actually see instead of something buried in the math. I'm building a dedicated bias-detection module that flags when outcomes start skewing along lines that have nothing to do with the job, so recruiters can catch it and fix it instead of scaling it without knowing. Being transparent isn't just about trust. It's about being able to spot what's going wrong while there's still time to do something about it.

The Solo Founder Reality

Working alone makes you honest about trade-offs.

I open-sourced some of the components, partly to invite people to poke holes in them, and partly because feedback from the community is about the closest thing a solo founder gets to a team of reviewers. I also leaned hard on talking to users. Early on, every single tester got a call from me so I could hear how it actually went for them. Those calls did more to shape the product than any roadmap I wrote.

It's slower than working with a team. There's no one to hand something off to at midnight. But there's a flip side. Every line of code is there for a reason, and I understand every decision in the system because I'm the one who made it.

What I've Learned

A few things caught me off guard:

• Transparency is something people will actually pay for. I figured buyers mostly wanted speed. What they really wanted was to be able to defend a decision, whether to a candidate, a manager, or a regulator.
• The hard part of AI isn't the AI. It's the judgment around it. What to automate, what to leave to people, and where the line between the two should be.
• Building alone is lonely, but it makes things clear. With no team to argue with, you can't hide behind a group decision. Every choice is yours, and so is every lesson.

What's Next

HCIA is live for mid-size and enterprise companies right now. Next up are deeper integrations with the big ATS platforms so it slots into the tools teams already use, plus a smarter bias-detection module.

The mission hasn't really changed though. Human-led hiring, powered by explainable AI. The AI does the heavy lifting, people make the call, and everyone can see why.

Let's Talk

If you're a founder or hiring manager who's sick of opaque tools that just ask you to trust them, give HCIA a try.

And if you've got feedback, especially the harsh kind, I want to hear it. Building on my own means I'm always learning, and most of the best lessons have come from people willing to tell me what's broken.