turbopuffer Serverless Vector and Full-Text Search

turbopuffer Inc. was founded in 2023 by Simon Hørup Eskildsen and Justine Li, who previously worked together for eight years on Shopify's infrastructure. It describes itself as a Canadian, remote-first company with offices in Ottawa, Montreal, New York, San Francisco and Paris, and is backed by Lachy Groom and Thrive.

turbopuffer is a search database built on object storage such as S3, with a memory and SSD cache in front, which lets it keep large datasets cheaply while serving hot data quickly. It provides vector search, BM25 full-text search, hybrid search and metadata filtering, with unlimited namespaces for multi-tenant applications. In production it hosts more than one trillion documents, handles over 10 million writes per second and serves 100,000+ queries per second for customers such as Anthropic, Notion, Atlassian, Linear and Cognition.

• Vector search: approximate nearest neighbour search with native filtering for high recall.
• Full-text search: BM25 engine (FTS v2) up to 20x faster than the previous version.
• Hybrid search: combines vector, full-text and numeric attributes for first-stage relevance.
• Unlimited namespaces: isolates each tenant or user dataset, with 250M+ namespaces seen in production.
• Object storage tiering: data on S3 with memory and SSD cache for hot namespaces.
• Integrated embedding: embedding moved inside turbopuffer to cut end-to-end search latency.

• Vietnamese SaaS products adding per-customer semantic and keyword search.
• Legal, banking and insurance firms building RAG over very large document archives.
• Developer tool companies offering code search across many repositories.

• Object-storage architecture that the vendor says is about 10x cheaper than alternatives.
• Vector search p50 of 14 ms on a 10M-document benchmark, with 90-100% recall@10.
• ANN v3 reaches 200 ms p99 latency over 100 billion vectors.
• Usage-based pricing from storage, writes and queries, with namespaces included.

AI coding tools, productivity apps and enterprise copilots index every user's documents, which demands cheap multi-tenant search at very large scale. SaaS and AI startups in Vietnam and Southeast Asia serving many customers face the same need to keep per-tenant search costs low.

turbopuffer Inc., founded 2023, a Canadian remote-first company with offices in Ottawa, Montreal, New York, San Francisco and Paris. Co-founders Simon Hørup Eskildsen (CEO) and Justine Li. Backed by Lachy Groom and Thrive.