What is a people search engine?
A people search engine is a search engine whose results are people rather than web pages. You describe who you're looking for, and it returns matching individuals - with their roles, companies, locations, and public work - assembled from records scattered across the open web.
What a people search engine does
A regular search engine answers “what pages mention this?”. A people search engine answers “who fits this?”. The unit of results is a person: their current role, employer, location, and the public work - repositories, talks, posts, funding announcements - that proves what they do.
That changes what a query looks like. Instead of keywords, you describe a person the way you’d describe them to a colleague: “founding engineers at seed-stage fintechs in London” or “ML researchers who published on retrieval and write Python”. The engine’s job is to turn that description into a shortlist of real, reachable people.
How does a people search engine work?
Under the hood, every people search engine solves the same three problems:
- Coverage. A person’s footprint is scattered: a professional profile here, a GitHub account there, a mention in a funding announcement somewhere else. The engine has to read across sources, not within one silo.
- Identity resolution. Those scattered records have to be merged into one person - the “J. Kowalski” who maintains a popular Rust crate and the “Jan Kowalski” who is a staff engineer in Berlin may or may not be the same human. Getting this wrong produces confident nonsense.
- Ranking. Thousands of people match “backend engineer” loosely. The engine has to score how well each person fits the whole query - seniority, location, stack, recency - and put the best matches first.
The traditional approach is a big pre-built index searched with filters. The newer, AI-driven approach adds a reasoning step: an agent interprets the query, plans sub-searches, reads sources live, and justifies each match. That’s the approach Refolk takes - here’s exactly how it works.
Natural-language search vs boolean filters
Classic sourcing tools make you speak their language: (“staff” OR “principal”) AND rust AND NOT recruiter, plus a stack of dropdown filters. That syntax is a skill, it’s brittle, and it caps what you can ask - a filter for “shipped something real in the last year” doesn’t exist.
Natural-language people search flips the contract: you say what you mean, and the engine does the translation. The practical differences:
- Expressiveness. “Engineers who left a unicorn in the past six months” is a valid query in plain English and nearly impossible as a boolean string.
- Recall. Booleans match literal keywords, so people who describe the same skill differently fall through. A model matches meaning.
- Iteration. Refining a boolean means rewriting the string. Refining a conversation means saying “narrower: only Berlin, only hands-on”.
Not a people-finder or background-check site
“People search” also gets used for consumer people-finder sites - the ones that sell addresses, phone numbers, relatives, and court records about a person you name. That is a different product with a different purpose. A professional people search engine is for discovery: you start from a role or a skill, not a name, and the output is someone’s public professional identity - not their private life. Refolk is the discovery kind. It reads the same public pages you could open in a browser, and nothing else.
What people use it for
- Recruiting and sourcing. The biggest use case: build a shortlist of qualified candidates in minutes instead of paging through filter results for an afternoon.
- Sales and partnerships. Find the actual decision-maker - “heads of data at Series B fintechs” - rather than a company switchboard.
- Investing. Track founders, early employees at breakout companies, and operators who just left one.
- Research. Map who is actually building in a space: maintainers, authors, and the teams behind the repositories that matter.
How Refolk does it
Refolk is an AI people search engine. You ask in plain English; I plan the search, read public LinkedIn and Crunchbase records, the public GitHub graph, and the open web live, and return a ranked shortlist with the evidence that earned each person their spot. Every search is inspectable - you watch it run - and real examples are published on the shared searches page. The full pipeline is on how it works.
Common questions
What is a people search engine?
A search engine whose results are people, not web pages. You describe who you're looking for and it returns matching individuals, with their roles, companies, and public work assembled from records across the open web.
How is a people search engine different from Google?
Google returns pages that mention your keywords and leaves the assembly to you. A people search engine resolves those mentions into individuals, merges each person's records into one profile, and ranks the people themselves against your query.
How is it different from LinkedIn or GitHub search?
Each of those searches one silo with rigid filters. A people search engine reads across sources - professional profiles, code, funding data, the open web - and lets you describe the person in a sentence instead of a boolean string.
Is a people search engine the same as a background check?
No. Background-check and people-finder sites sell personal records - addresses, phone numbers, court filings - about a person you already know. A professional people search engine like Refolk works the other way: you describe a kind of person, and it discovers who fits, using only their public professional footprint.
Where does the data come from?
Public sources: public LinkedIn and Crunchbase records, the public GitHub graph, and the open web. The same pages you could open in a browser, read and cross-referenced at search time.
Can AI find people for me?
Yes - that's what an AI people search engine does. Ask Refolk for "staff engineers in Berlin who maintain popular Rust crates" and I plan the search, read the sources live, and return a ranked shortlist with the evidence behind each match.