Candidate search across 800 million profiles in 1 to 2 seconds
- candidate records searched
- 800M+
- query response time
- 1–2 s
Context
Whitecrow Research is an HR Tech company building a recruiter search product that competes with LinkedIn Recruiter. Recruiters search a pool of more than 800 million candidate records, so search speed and result quality are the product.
Rewriting the search APIs
The legacy Elasticsearch backend APIs were slow at this scale. I led the rewrite. Queries across the full 800M+ record set now return in 1 to 2 seconds, which made the platform much faster for recruiters to use.
Ranking by what the recruiter asks for
I designed Elasticsearch queries that rank candidates by each recruiter’s own preferences, not by one fixed relevance order. Matches got more accurate and recruiters spent less time sifting through results.
AI-driven job description parser
I built both halves of a module that parses job descriptions with AI and matches them against the candidate pool:
- Backend APIs in Spring Boot (Java) and Node.js.
- Frontend in Angular.
A recruiter can paste in a job description and get matching candidates without writing a search by hand.