TrueHire — Résumé credibility checker

2025Applied ML

An LLM desktop app that checks a candidate's résumé against their GitHub and scores how credible it is.

TrueHire: résumé, LinkedIn and GitHub in, Gemini credibility score out
Google GeminiPrompt EngineeringElectronLit

Problem

Recruiters see inflated résumés, borrowed projects and timelines that don't line up, and real background checks are slow and expensive. We wanted a first pass that takes seconds.

What I built

Built with a teammate for a hackathon.

  • Inputs: a résumé (PDF, DOC, DOCX or TXT), a LinkedIn export and a GitHub username.
  • AI pipeline: files are parsed and normalised, then sent to Google Gemini with prompts that ask it to check every claim against public GitHub activity.
  • Output: a structured credibility score with its reasoning, plus red flags such as suspicious projects, unsupported claims and timeline gaps.
  • App: an Electron desktop UI (Lit components) that streams Gemini's analysis to the recruiter as it arrives.

What I learned

Most of the work was prompt design: getting the model to cite the evidence for each claim and return a fixed structure the UI can render, instead of a confident paragraph.