Slide Hub
An internal document search and reuse tool at Databricks for field engineers. Its core capability is paragraph-level (chunk/passage) vector retrieval for quickly locating reusable technical solutions and presentation materials. At a scale of four million passages, it employs a hierarchical retrieval architecture to balance recall and latency.
Timeline (last 90 days)
The Databricks team went through the whole process from prototype validation to production deployment when developing the internal application Slide Hub, which serves field engineers for quickly finding and reusing internal documentation materials.
Slide Hub's document library grew to the scale of 4 million paragraphs.
When the number of paragraphs reaches 4 million, simple full vector scanning can no longer meet the latency requirements of interactive search.
All Facts (3)
The Databricks team went through the whole process from prototype validation to production deployment when developing the internal application Slide Hub, which serves field engineers for quickly finding and reusing internal documentation materials.
50%UnverifiedSlide Hub's document library grew to the scale of 4 million paragraphs.
50%UnverifiedWhen the number of paragraphs reaches 4 million, simple full vector scanning can no longer meet the latency requirements of interactive search.
50%