Veritas Argus Toth — an AI screening assistant for recruiters
An interview-call recorder plus a backend that turns the recording, resume and E-Staff data into a finished candidate report.
The task
The agency's recruiters ran interviews in the browser and then manually listened back to the recordings to write a per-vacancy conclusion for each candidate — hours of work per person. Candidate and vacancy data lived in the corporate E-Staff system, the call transcript was never saved anywhere, and the report format had to be rebuilt by hand each time to match a specific client's or vacancy's requirements.
The solution
We built a Windows desktop app that records the browser tab's system audio via ffmpeg loopback during the call and uploads it to the backend. The backend transcribes the conversation with Google Speech-to-Text (aware of language and speaker count), pulls the candidate's resume and the vacancy card straight from the E-Staff API, and feeds all of it — transcript, resume, vacancy, prior analysis — into a single prompt for Gemini or Claude. The resulting report is rendered from a template configurable per vacancy or client in the admin panel, then converted to PDF via LibreOffice with no manual formatting.
The result
- The path from call to finished candidate PDF report is fully automated — no manual listening required
- Report templates per vacancy and client are edited in the admin panel without touching code
- Direct E-Staff sync removed duplicate manual entry of candidate and vacancy data
- Reports factor in the candidate's prior analysis context, not just the latest call