Your best applicant has a bad resume. Keyword filters lose them. Hirelight interviews them.
You're a 6-person startup with 200 applicants. In that pile: a career changer, a Next.js dev applying to your "React" role, a great engineer who writes terrible resumes. Hirelight screens for real skills instead of keywords — and can interview every single applicant, so answers decide who you meet, not formatting. You make the call. Everyone hears back.
Priya Raman
Platform Engineer · applied 2h ago
Kubernetes "built an internal developer platform serving 40 teams"
React "5 years of Next.js — same ecosystem, counted"
Go "mostly Python; one Go service in production"
Interview: 82/100
Integrity: clean · 0 flags
From "we should hire someone" to a shortlist, without the 3-week detour.
Post a job in 5 minutes
Describe the role, list your must-have skills, say what a great candidate looks like. You get a public job page and a structured interview, generated and fully editable.
Every resume gets read
Not keyword-matched. Read. Each must-have skill gets a verdict with a quote from the resume as proof — and adjacent experience counts: Next.js is React evidence, a career changer's real projects aren't invisible.
Structured written interview
Same questions for everyone, so it's fair and comparable. Turn on interview-first mode and every applicant gets one automatically — no resume gate. Timing and typing patterns quietly flag copy-pasted AI answers. No webcam.
You decide. They hear back.
Ranked candidates, full transcripts, flagged evidence, one review screen. You click the button. Rejected candidates get honest, constructive feedback automatically.
Things we refuse to build. Even if you ask nicely.
Most hiring tools compete on features. We also compete on the features we won't ship, because trust is the actual product.
Webcam proctoring
Staring at candidates through a camera is surveillance, not assessment. Behavioral signals catch cheating without the creep factor.
Auto-rejection
The AI never rejects anyone. It ranks and shows evidence. A human clicks the button, and your name is on the decision.
Black-box scores
A number with no reasoning is a liability, not a shortcut. Every score comes with the exact evidence behind it.
Keyword bingo
A teacher who built the school's scheduling system can be a great junior engineer. A decade of Next.js is React experience. Systems that match strings will never see either — ours reads for the capability and shows its reasoning.
Ghosting
The default in hiring is silence. Here it's impossible: every applicant gets a definitive outcome, with feedback when it's a no.
What we build instead
Evidence per skill, quoted from the resume. Structured interviews scored per answer. Behavioral integrity flags with severity. Feedback for every rejection. Boring? Maybe. It's also why candidates finish your process instead of abandoning it.
Candidates are users too. Radical, we know.
Before you apply, the job page tells you exactly how the process works and what the AI does. After you apply, you get a tracking link that updates at every step. And whatever happens, you get an answer. If it's a no, you get real feedback, not "we went with other candidates" copy-pasted into the void.
Track my application →Your application · Platform Engineer
No ghosting, promised. This page updates at every step.
Pricing: free.
We're in pilot. You get the whole product, we get your complaints. When pricing exists, pilot teams keep a grandfathered deal. There is no credit card field anywhere in this product. We checked.
Claim a pilot spot →Questions people actually ask
→Is this legal? AI hiring sounds regulated.
It's regulated when AI makes decisions or scans faces. Hirelight does neither: no biometrics, no auto-rejections, full disclosure to candidates before they apply, and a human making every call. That's the compliant end of the spectrum, by design.
→Won't candidates just answer with ChatGPT?
Some will try. Response timing, focus switches, and paste patterns get flagged with severity levels, and the questions favor debugging and judgment over things a model answers cold. You see the flags next to the answers and judge for yourself.
→What does it cost?
Nothing right now. We're in pilot and want ten opinionated small teams using it hard. When we charge later, pilot teams get grandfathered pricing. No credit card exists anywhere in the product.
→Our must-have is React. What about the person who only writes Next.js?
They get credit. Screening treats skills as capabilities, not strings: Next.js is React evidence, Fastify counts toward Express, ten years in an ecosystem beats a missing keyword. And if you switch a job to interview-first mode, every applicant gets the written interview automatically — candidates who bomb the resume but ace the answers get flagged as exactly that, instead of disappearing.
→We already have a process. Sort of.
If "sort of" means a shared inbox and a spreadsheet, that's exactly who this is for. Keep your process, drop the spreadsheet: post the job, share one link, and review ranked candidates with evidence.
Your next great hire is in that pile of 200 resumes.
Possibly attached to the worst resume in it. Don't let a keyword filter make that call for you.
Post your first job free →5 minutes to set up. First shortlist in a day.