How Inferr verifies an AI engineer (proof of work, not résumés)

Every hiring platform claims its candidates are “vetted.” Almost none can tell you what that means. Here is exactly how Inferr verifies an AI engineer — in plain English, with nothing taken on faith.

The problem: every signal is self-reported or gamed

A résumé is a list of claims. A green commit grid is trivially faked. A take-home test measures who has a free weekend. None of them tell you whether an engineer has actually shipped AI systems that work. Inferr's job is to replace claims with corroborated evidence.

Step 1 — a local scan that never reads your code

It starts with one command: npx inferr-scan. It runs entirely on the engineer's machine and reads git metadata only — commit counts and dates, language and framework detection, and AI-tool fingerprints. It never reads file contents, file paths, or credentials. The scanner is open source, so anyone can audit exactly what it collects before running it.

• What it reads: commit cadence, languages, frameworks (LangChain, vLLM, PyTorch…), AI-tool usage.
• What it never touches: your source code, file contents, real paths, secrets.
• What leaves your machine: aggregate signals only — never the work itself.

Step 2 — public corroboration of real production work

A scan proves you write code. It doesn't prove anything reached the world. So engineers attach shipped work, and Inferr checks each claim against public signals only:

• A live product — is the deploy actually reachable?
• A package — real downloads on npm, PyPI, or Hugging Face?
• Open source — merged PRs in real repositories, weighted by the repo's standing?

Step 3 — a Proof-of-Work score and verification tiers

Those signals roll up into a production-weighted Proof-of-Work score and a verification status. A profile is unverified until it has at least one corroborated production achievement; it becomes verified with real shipped work, and elite with sustained, independently-confirmed output. Employers rank on that score — not on how well someone writes a bullet point.

What this means for both sides

Engineers get credit for what they actually built, and a faster path to remote work without polishing a résumé. Employers see a candidate pool where every profile has already cleared the bar — so the first conversation is about fit, not about whether the claims are real. Proof, not promises.