Build a Recruiting CRM with AI

Yes. Wire up the OpenAI or Anthropic integration and pass uploaded resume PDFs through the model with a structured-output prompt. You get back JSON with name, contact info, current role, work history, education, and skills, ready to populate your candidate record. Cost is roughly $0.01 per resume on Claude Haiku, much cheaper than the parsers built into commercial ATS tools.

Connect a calendar via Google or Microsoft OAuth. The CRM reads each interviewer's availability, lets you propose 2-3 slots to candidates, and creates calendar invites with video links automatically. You can also build a candidate-facing self-scheduling page (Calendly-style) tied to specific interviewer pools.

Yes. Generate a per-client portal as a separate route in the same app. Clients log in with magic-link auth and see only the candidates you've submitted to them. They can leave feedback or rate candidates without seeing the rest of your pipeline. Webtwizz scaffolds the row-level security so candidates from one client never leak to another.

Because you own the database (Postgres on Supabase), candidate deletion is a real DELETE, not a soft-delete behind a vendor's API. Add a "Delete candidate data" button that triggers a cascading delete across candidates, notes, scorecards, and interviews. You can also schedule automatic deletion of rejected-candidate records after N days, which most commercial ATS tools charge extra for.

A working version, candidates table, pipeline view, notes, basic scheduling, takes about 2 hours of prompting and visual tweaking. Production-ready (auth, calendar sync, resume parsing, client portals, scorecards) is a weekend. Compared to integrating Greenhouse and customizing it to your process, it's a similar timeline at a fraction of the ongoing cost.