Case Study: Pylon achieves scalable multi-tenant semantic search with turbopuffer

PYLON ships multi-tenant semantic search in 3 days with Turbopuffer

Pylon, a B2B customer support platform, needed to implement a multi-tenant semantic search feature for its AI tools but faced significant challenges with OpenSearch. The API was counterintuitive and its architecture led to prohibitive costs that scaled linearly with their number of customers. They turned to Turbopuffer for its semantic search engine to solve these issues.

By implementing Turbopuffer, Pylon shipped its semantic search feature to production in just three days. The solution provided a simple API and an object storage-native architecture that kept costs proportional to usage rather than tenant count. This resulted in indexing over 300 million documents with a p90 query latency of 24ms, enabling Pylon to run more experiments and index more data without cost anxiety.

PYLON