Technical Talks

Simon Eskildsen

Co-Founder | Turbopuffer

Mickey Liu

Software Engineer | Notion

Go Back

Billion-Scale Vector Search on Object Storage

ABOUT THE TALK

Data Eng & Infrastructure

Vector Search at Scale: How Notion Built Billion-Vector Search Infrastructure | Explore the architecture behind Notion's enterprise-scale vector search system, powering one of the largest semantic search implementations in production. Learn advanced techniques in embedding pipeline design, distributed vector processing, and optimal storage strategies using Spark and Turbopuffer. This technical deep-dive covers LSM indexing, RAG (Retrieval Augmented Generation) implementation, and practical approaches to query optimization. Discover battle-tested strategies for building and scaling production-ready vector search systems capable of handling billions of vectors with high performance and reliability.

Co-Founder

Simon Eskildsen

Turbopuffer

Simon Hørup Eskildsen is co-founder of turbopuffer, a fast search engine on object storage. Previously, Simon was a principal engineer at Shopify, scaling the system from 1K to 1M requests per second. When time allows, he writes about napkin math on his or the turbopuffer blog and maintains a popular reference on the topic on Github: sirupsen/napkin-math.

Software Engineer

Mickey Liu

Notion

Mickey Liu is a software engineer on the AI & Data Platform team at Notion, helping build and scale data needs for Notion AI. Before Notion, Mickey worked on the ML Platform team at Twitter building the Feature Management system to enable ML model training and serving use cases for ads and recommendations.

Technical Talks

Billion-Scale Vector Search on Object Storage

ABOUT THE TALK Data Eng & Infrastructure

ABOUT THE TALK

Data Eng & Infrastructure