Since DuckDB’s inception in 2018, the tool has grown rapidly in the hearts of developers because of its versatility, ease of use and speed. Due to DuckDB’s in-process nature and concurrent write/read constraints on its own storage format, there is a widespread belief that DuckDB is the tool that you quickly spin up in your laptop for a quick local analysis. However, this is a very narrow way of looking at the possibilities that DuckDB can offer. This talk aims to challenge that misconception by exploring the full breadth of what DuckDB can do in collaborative, production environments.

I will introduce DuckDB and its versatility as a database and a query engine, focusing on its multi-player capabilities. DuckDB can now query data anywhere from lakehouse formats like Delta or Iceberg in remote object storage to any database that you may think of via ADBC, ODBC, JDBC or custom extensions. At DuckDB Labs, we’ve also developed DuckLake, our own lakehouse format, which aims at solving some of the shortcomings of previous lakehouse formats (i.e. metadata overhead and concurrency model). At the end of the talk I will disclose a new upcoming product coming out in May that will open a new realm of possibilities for the DuckDB multi-player ecosystem.

Guillermo Sánchez Dionis works in Developer Relations and Product at DuckDB Labs, where he has been contributing to the DuckDB ecosystem since July 2025. Before joining DuckDB Labs, he spent eight years as a Data and Software Engineer, starting in the era of migrating Hadoop clusters to cloud platforms like Snowflake, Databricks, and BigQuery. Over time, he grew increasingly skeptical of the complexity and cost that these platforms introduced, which led him to DuckDB — a tool he fell in love with for its simplicity and its philosophy of building things the right way.
Originally from Madrid, Guillermo now calls Amsterdam home. When he's not thinking about query engines, you'll find him climbing, exploring nature, or in search of a good meal. He also firmly believes that not every data talk needs to mention AI — and this one won't.