Case Study
Evidence-based talent marketplace
Production stack in 6 days: 30 tables, ~470 tests
Custom production app · Recruiting
The goal
Sell recruiters a subscription to searchable, cited evidence of what candidates have actually done, while candidates keep control of what gets published.
The problem
A resume compresses years of work into a page of claims. Recruiters can't see the work behind those claims, and candidates have no say in how their work gets pitched.
What we did
Candidates upload the documents they already have: PDF, Word or Excel. The app extracts the text privately, drafts a long-form account of the work with AI, suggests what to anonymize, and publishes only what the candidate approves, as a fixed snapshot they can update or take down.
Recruiters search by role, and every answer cites the section it came from. Their AI agents get the same read-only search through an MCP server, with hashed credentials and a log of every access. Row-level security in Postgres, fail-closed defaults and deletion receipts were in place before the first user.
The results
- ✦136 commits from spec to working product between September 7 and 12, 2026
- ✦14 database migrations and 30 tables, with row-level security in the database
- ✦About 470 automated tests plus a browser acceptance suite, run in CI on every change
- ✦Deployed from Docker with health checks and automatic restarts. Billing is the last piece before launch.
What it does for revenue
- Margins
- Recruiters pay a flat $80 a month, and the seat covers their authorized AI agents. The revenue is recurring from the first customer, with no placement fees to chase.
- Leads
- Every candidate who shares a profile brings a recruiter in free. Reading costs nothing, saving a profile needs a free account, and wider search is the paid tier.
- Visibility
- Published profiles are searchable by recruiters' AI agents over MCP, so candidates reach automated screening as well as human readers.
- Speed to market
- Six days from spec to a tested, deployed stack, so the business can start testing pricing with recruiters sooner.