Candid Consolidates Six Query Engines Into One, Enabling AI Without Expanding Headcount

About

Candid is a nonprofit that connects nonprofits and funders, providing the data each needs to do good. With a small engineering team and a data-intensive mission, Candid needed infrastructure that could scale without requiring a large team to operate it and support both analyst workflows and emerging AI capabilities.

Challenge

Candid’s data environment had grown into a sprawl of six or more query engines: SQL Server, DynamoDB, Elasticsearch, and others. Analysts were switching between multiple tools and losing significant time to cross-system queries.

An attempt to consolidate onto Snowflake made things worse: centralizing data meant making copies that were expensive to maintain and quickly went stale. Meanwhile, the data science team building in-house AI systems hit the same wall: AI agents can’t open multiple programs like analysts can. They need everything accessible in one place.

Two teams had independently discovered Trino as the answer. When they realized they were solving the same problem with the same tool, the case for a shared, managed solution was clear.

Solution

Candid chose Starburst Galaxy over self-managed Trino because, as a nonprofit, they didn’t have the resources to run Trino infrastructure themselves. Starburst Galaxy handles cluster management entirely, freeing the team to focus on data products.

Rather than centralizing data, Candid federates — leaving SQL Server, DynamoDB, and Elasticsearch where they belong and letting each engine do what it does best. For data that benefits from lakehouse price-performance, Galaxy’s managed ingestion brings it into Apache Iceberg natively, including Candid’s JSON-generating systems.

The same foundation now powers Candid’s first production AI system: an internal research tool that lets academic partners and institutions query Candid’s data conversationally.

“It allows us to federate our data with one focal point, and it manages all the underlying infrastructure. An organization of our size is able to utilize all of those features without having to have a large team to back it up.” — Charlton Julius, Senior Data Architect

Results

Starburst Galaxy has become the foundation of Candid’s data platform, delivering impact across engineering efficiency, AI enablement, and cost:

• Six-plus query engines consolidated into one — analysts have a single access point for all data infrastructure, eliminating multi-tool overhead
• Saved the equivalent of 5–10 hires — centralized access control, data products, and cluster management in one place eliminates the infrastructure overhead that would otherwise require a significantly larger team
• Zero stale data for AI agents — before Starburst, AI systems queried copies of data that degraded over time; agents now hit live source systems directly through Starburst
• Faster iteration — with one place for everything, the team iterates on data products significantly faster than when managing six separate engines
• First AI product in production — Candid’s internal research tool, enabling conversational access to Candid’s data for academic and institutional partners, launched on the Starburst-backed foundation

As Candid continues to expand its AI capabilities, Starburst Galaxy remains the connective layer — for analysts, for agents, and for any new system that needs access to Candid’s data.

“Starburst is kind of our foundation right now. It’s become foundational to everything we’re doing. Even if AI goes away, we’re still going to be using the Starburst platform for everything that we’re doing.” — Charlton Julius, Senior Data Architect, Candid

More resources: Performant Analytics Engine – Charlton Julius | Starburst AI & Datanova