About
ParadeDB builds a Postgres-native search and analytics engine: its pg_search extension adds full-text, BM25, vector search, filtering, and aggregations inside the system of record. It targets teams already using Postgres, from startups to enterprises such as Alibaba Cloud, Bilt Rewards, and Modern Treasury, and differentiates by combining transactional database guarantees and standard SQL with search-engine capabilities without requiring a separately synchronized search system.
Market
ParadeDB competes in the database-embedded search, vector retrieval, and real-time analytics market, positioning itself as a way to keep application data, search indexes, vectors, and aggregations in one Postgres system instead of maintaining a separate search cluster and denormalized data copy. It differentiates through a Postgres extension built on Tantivy that provides Elastic-like BM25 search, fuzzy matching, faceting, hybrid search, and SQL/JOIN compatibility with zero ETL, while retaining Postgres transactions, backup, and high-availability capabilities.
ParadeDB is aimed at software and data engineering teams—especially in SaaS, finance/fintech, retail, and other data-intensive businesses—whose primary database is managed or self-managed Postgres but whose search workloads have outgrown native tsvector/pgvector capabilities. Its customer range spans fast-growing startups to large enterprises, with likely buyers including database, platform, and search engineering teams.
At a Glance
Problem
ParadeDB addresses the “two systems” problem for teams whose application data already lives in PostgreSQL but requires Elastic-quality search. PostgreSQL is strong for reliable transactions, yet its native search capabilities can become inadequate on large datasets or complex queries; moving search to Elasticsearch or another dedicated engine introduces a separate system, synchronization pipelines, operational overhead, and potential data inconsistencies. The core pain is therefore both technical and economic: teams must spend engineering time maintaining ETL and duplicated indexes, or accept slow, inaccurate search.
The killer use case is search-heavy Postgres applications such as payments, financial records, marketplaces, and large document collections, where users need full-text search, fuzzy matching, filtering, joins, faceting, and aggregation over millions or billions of rows. Bilt’s Postgres search produced inaccurate results, high latency, and frequent timeouts, forcing engineers into multiple API calls and client-side result assembly; after adopting ParadeDB, it reduced query timeouts by 95%. Modern Treasury faced similar complexity across datasets exceeding 100 million records, where the permutations of filters and query plans overwhelmed traditional indexes.
Product / Service
ParadeDB is a Postgres-native search and analytics engine delivered primarily through the open-source pg_search extension. It adds BM25-ranked full-text search, Boolean and fuzzy queries, faceting, hybrid search, vector retrieval, and aggregations while keeping application data and search indexes in PostgreSQL. Queries remain SQL, support normal PostgreSQL joins, and can update the search index in the same transaction as writes, eliminating the external synchronization pipeline. Under the hood, its custom ParadeDB index is built on Tantivy, the Rust implementation of Lucene, giving it dedicated-search-engine performance while preserving PostgreSQL’s transactional model.
The delivery model spans AGPL-licensed Community software, self-hosted deployments inside PostgreSQL or Kubernetes, Enterprise deployments with high availability, read replicas, support, and SLAs, plus BYOC and fully managed cloud offerings. This lets a team adopt the technology without introducing a new query language or search cluster, reuse its PostgreSQL expertise and infrastructure, and reduce the operational and migration burden of building search separately. Customer results illustrate the benefit: Bilt needed only minor SQL changes and Modern Treasury deployed ParadeDB as a logical replica across seven clusters without new data pipelines or sync jobs.
Market
ParadeDB competes in database and search infrastructure, positioning itself as an open-source, Postgres-native alternative to Elasticsearch and as a way to consolidate application data, search, vector retrieval, and analytics in one database. Its broader competitive set includes dedicated and hosted search products such as Algolia, Meilisearch, and Typesense, as well as PostgreSQL’s built-in full-text and trigram capabilities. The differentiation is not simply search relevance; it is search quality combined with SQL, joins, ACID guarantees, PostgreSQL compatibility, and less operational complexity for teams already standardized on Postgres.
The company has substantial early traction rather than looking like a pre-revenue prototype. ParadeDB launched from Y Combinator’s Summer 2023 batch and reports production use since December 2023; its July 2025 Series A announcement disclosed $12 million led by Craft Ventures with participation from existing investors including Y Combinator. The same announcement reported more than 7,000 GitHub stars and 100,000 installs, while the company cites enterprise users including Alibaba Cloud, Bilt Rewards, Modern Treasury, and other Fortune 1000 companies; its customer portfolio expanded with 2026 case studies for Modern Treasury, Terrapin Finance, and Cofactr. The reviewed public materials do not disclose revenue or ARR, so revenue scale cannot be established, but customer deployments, open-source adoption, and enterprise financing demonstrate commercial market validation.
Founders & Leadership
Funding History
Y Combinator (existing investor; lead not disclosed)
Craft Ventures
Recent News
ParadeDB examines why the same Top-K query can produce different results under different workloads or scheduling conditions, comparing benchmark behavior with Postgres full-text search.
Render added ParadeDB as an official integration, enabling one-click deployment of ParadeDB onto Render.
Railway added ParadeDB as an official integration, allowing users to deploy a ParadeDB instance with one click and get running in under a minute.
ParadeDB explains how search-engine principles can improve Top-K performance in Postgres, including the use of a compound index designed for varied filtering, sorting, and text-search workloads.
ParadeDB presents a production-grade personalized search engine built entirely within Postgres using BM25 retrieval and vector-based reranking, without external infrastructure.
ParadeDB introduced faceted search for PostgreSQL, bringing search-engine-style faceting and single-query aggregations to its Postgres-native search system.
Sacra reported that Neon’s integration of ParadeDB’s pg_search brings Elasticsearch-grade search capabilities directly into Postgres and helps Neon address cloud search.
ParadeDB detailed a v2 API that eliminates schema duplication, simplifies tokenization, and provides transparent search operators for intuitive SQL-based full-text search.
ParadeDB 0.20.0 added fast search aggregations, made the v2 API the default experience, and improved write performance, strengthening ParadeDB as an Elasticsearch alternative inside PostgreSQL.
Tinybird’s comparison coverage described ParadeDB as a newer option with fewer large-scale production deployments and noted that its search extension had reached stable release v0.16.4 in July 2025.
Active Roles
2Business Model
ParadeDB monetizes through a free, open-source Community edition licensed under AGPL-3.0 and Enterprise access sold through contact sales with custom pricing, adding closed-source enterprise features and production support. It also provides managed deployment options, including BYOC within customers’ cloud environments and a fully managed cloud offering.