About SafeHire

Job hunting is stressful enough
without wondering if the offer is real.

SafeHire is an AI investigator that sits between a job seeker and a job posting — reading it the way a skeptical friend with a cybersecurity background would, then telling you exactly what it found and why it matters.

How it started

Built for the Build with Gemma Hackathon 2026.

SafeHire began as a hackathon build with a narrow, practical question: could an open model read a job posting the way a careful human would, and catch the same red flags — fake companies, cloned recruiter profiles, upfront "training fee" requests — before an applicant ever hands over money or personal information?

Job scams have gotten harder to spot on sight. Postings borrow real logos, recruiters message from convincing LinkedIn profiles, and pressure tactics ("we need an answer today") are designed to short-circuit exactly the kind of careful check most people don't have time to run. SafeHire runs that check for them, in seconds.

Under the hood, Gemma 4 handles the reasoning layer — reading extracted text, weighing it against known scam patterns, and writing the plain-English explanation behind every risk score — while a supporting pipeline handles OCR, entity extraction, and cross-checks against domain, email, and recruiter data.

Hackathon Build

Build with Gemma
Hackathon 2026

  • ModelGemma 4
  • CategoryReal-World Safety
  • StackPython · FastAPI · Gemma 4
  • StatusActive Development

“Nobody should lose money, time, or personal data to a job that was never real in the first place — and nobody should need a security background to find that out.”

Our mission

What we build around.

Four principles guide every feature, from the OCR step to the final report.

Privacy first

Whatever you paste or upload is used to generate your report and nothing else. Your data is encrypted and never shared or sold.

Explainable, not just a score

A number alone isn't a reason to trust or distrust an offer. Every report shows the specific evidence — a flagged domain, a mismatched name — behind the verdict.

Human in the loop

SafeHire flags risk; it doesn't make the final call for you. The goal is a clearer decision, not a replaced one.

Built in the open

SafeHire is an open, evolving hackathon project. The architecture, prompts, and detection logic are visible rather than a black box.

Got a job posting that feels off?

Paste it in and see what SafeHire finds — it takes less time than reading this page did.

Start Investigation See How It Works