From attendee list to live search,
in a weekend.

1. 01
   Attendee list

   Scraped Brella

   Started with the public attendee directory on Brella — ~2,000 names with roles, companies, and (for some) a LinkedIn URL.

2. 02
   LinkedIn — pass 1

   Scraped LinkedIn URLs

   Where Brella had a LinkedIn URL, I ran it through an Apify scraper to pull the canonical profile — headline, about, experience, location.

3. 03
   LinkedIn — pass 2

   LinkedIn search

   For everyone else, two queries per attendee — “Name + Sydney” and “Name + Company” — kept the top-confidence match. ~95% hit rate; the remaining ~5% had no clean match and stayed Brella-only.

4. 04
   Storage

   Local SQLite

   Loaded the merged dataset into a local SQLite database. File-based, fast, and queryable straight from Node — no servers, no migrations.

5. 05
   Search

   Hybrid RAG + rerank

   Your prompt fans out to keyword and semantic retrieval, then an LLM reranks the top candidates against your intent. That ordered list is what you see.