AI Recruitment Pipeline
an interviewer that actually listens. it scores you against the job's own rubric, then decides whether what you just said needs digging into — and messages you on telegram to ask. the hard part wasn't making it ask questions. it was teaching it when to stop.
The screening layer of Zemenay's recruitment automation: a client's job request runs to a hired candidate with almost no manual steps. Gemini scores each applicant against that opening's rubric and returns a structured verdict — a 0-100 fit score, a separate score judging the CV document itself, concerns, red flags, and up to three gap questions. The gap questions are the interesting part: they fire only when a must-have is unverified, never re-ask something already answered, and go out over a Telegram agent loop where the candidate can reply with photos or PDFs that feed straight back into the next scoring pass. Inngest runs the orchestration. The model is explicitly told it never makes the final call — a human recruiter does. Cut time-to-hire by 40%.