Case study · Product
A referral network nobody could search became 80,000 people you can.
- Network
- 81,831 people
- Connections
- 90,972
- Built in
- One all-hands
- Discipline
- Talent Engineering
The Challenge
The referral pool was whatever the recruiting team happened to know.
About 10,000 connections, all of them inside a handful of recruiters' heads and inboxes. Everyone else in the company had a network nobody could see, let alone search. Asking for referrals meant a Slack message that described the role and hoped somebody recognised it.
The usual fix is to ask everyone to export their connections one by one. Nobody does it, so the data never arrives.
What We Built
Ingest once, enrich everything, then search it like a person.
01
Ingest
A browser extension collects each person's connections. We ran it once during an all-hands: 400,000 connections in a single session, with staged writes so the spike couldn't drop data.
02
Enrich
A raw connection is just a name and a URL. Each profile is enriched with work history and context, then deduplicated: 400,000 raw connections resolve to about 80,000 distinct people.
03
Search
You describe the person you're looking for in plain language instead of guessing keywords, and results come back ranked with the reasoning attached.
04
Route
Every result shows who on the team can make the introduction, and whether the person is already in the ATS, so nobody re-sources someone the team spoke to last quarter.
81,831
People in network
90,972
Connections
51
Contributors
100%
Enriched
Search the network
Recruiting Coordinator · Northwind Labs
Exact title at a large AI lab, has run scheduling at volume, and brings robotics-adjacent recruiting history.
Senior Recruiting Coordinator · Halcyon Robotics
Three years coordinating hardware and research loops, owns interviewer training, ran a scheduling migration end to end.
Recruiting Operations Associate · Brightpath
Ops-leaning coordinator: built the reporting the team runs on and is the de facto ATS admin.
Illustrative. Real deployments run against the company's own network. The people shown here are invented.
10K → 80K
Searchable potential referrals, from one all-hands
400K
Raw connections ingested before enrichment and dedupe
Semantic
Search by describing the person, not matching titles
1 session
Time it took to build the network
Why It Sticks
A list decays. A tool gets used.
A one-off export is stale the week after it's taken. This runs as a product: new joiners add their network, feedback on a result is remembered against the role, and the ATS check means the team stops re-sourcing people it already spoke to.
It's also documented and handed off. The recruiting team runs it without us, which is the only version of this worth building.
Find out where your team is losing output.
A short diagnostic on your funnel, your data, and where the hours are going. No pitch deck.