Querying billions of rows of LinkedIn data in plain English for Blueprint
Blueprint finds outbound prospects in a LinkedIn database of billions of rows. BetterBrain built models that turn plain-English requests into SQL and learn from each query, helping drive over $600,000 in revenue in six months.
Revenue driven in six months
BetterBrain brings together expertise from top institutions and innovative companies, combining deep academic AI knowledge with practical industry experience.
About the client
Blueprint is a US outbound agency. It runs outbound campaigns for its clients and sends hundreds of thousands of emails a month.
The challenge
Every campaign starts with a list of the companies and people that fit the client’s ideal customer profile. Blueprint pulled those lists from a LinkedIn database with billions of rows, and each new profile meant someone writing a new SQL query by hand.
The queries took time to write and needed someone who knew both SQL and the data, and that slowed every engagement down.
The solution
BetterBrain built models that write and run SQL from a plain-English request, such as the kind of company and role a campaign is after. Query generation runs on OpenAI’s GPT-4o.
The system learns from its own history. It keeps every request alongside the query that answered it, and when a new request comes in it finds the closest past examples with keyword search (BM25) and vector embeddings. Each answered query makes the next one easier to get right.
Impact
$600,000+ in revenue in six months
Blueprint’s team could query its data and finish engagements much faster, which drove over $600,000 in revenue in six months.
No SQL required
Building a prospect list now starts with a sentence instead of a query.
