DuckCon #7 - Amsterdam (Event Notes)

Conference about DuckDB.

When: 2026-06-24

DuckCon #7 in Amsterdam – DuckDB
YouTube: DuckDB Live Stream

# Notes and Slides

Find full presentation titles and speakers in Agenda header at the bottom.

# DuckDB 2.0

Mark Raasveldt talks about 2.0 version.

# Object Stores

# Async I/O

async IO for Object store to make parquet on S3 (and other formats) much faster (on network storage)

# C API

Extensions are done against the internal C API, which means you need to choose a version.

Goal to make an API that stays stable and more reliable.

# Qustions

How do you feel about major industry leaders like databricks building next gen engines (Reyden) that are modeled off of duckdb?

They are flattered about it 😀.

# SQLFrame

SQLFrame:

Based on SQLGlot

dbt state is using SQLGlot (which was SQLMesh before) is using it too.

Ryan Eakman - also employed by Fivetran now:

SQLFrame used to post PySpark to DuckDB. Great for Apache Spark Alternatives.

Speed improvements

Why it worked for us:

# Phages and DuckDB: The Mighty Duo against the Giants

Phages and DuckDB: The mighty duo against the giants

Helped them become faster for biology and genomics analysis:

Great overview and solutions with Duckdb - e.g., from Python Loops to DuckDB:

It changed the analysis from files to full folders.

# Ggsql

ggsql

Grammar of Graphics

Why not Power BI - because code is better than point-and-click and not close-source:

He says visualizations should be expressed in code. I’d say Code Is Still the Best Abstraction.

SQL is declarative and with grammar of graphics, it gives you ggsql.

# Spotify Using DuckDB

Using a lot of Materializations.

DuckDB reduced the load by a lot:

# Grep Lakehouse with DuckDB (and DuckLake)

Extending DuckLake with semantic search:

Parquet Sidecars:

# MariaDB + DuckDB (as a Storage engine)

Speed improvements:

Comparison against ClickHouse and MariaDB:

# Lightning Talks

# Native App Development Using DuckDB

Having a Data Engineering Toolkit that is hybrid with Tauri, DuckDB and Microsoft Fabric:

OneLake connection upload via Azure Blob Storage:

most important are Embedded OLAP (one of the key DuckDB Use Cases), encryption, and fabric suppport:

# pg_lake

Snowflake with pg_lake:

Direct inserts via DuckDB:

The architecture with avoiding adding DuckDB to every Postgres to not have to create a connection everytime - that’s why they’ve build a Sidecar (same as above talks grep lakehouse (semantic search)).

the iceberg logic lives incide Postgres (they rewrote the whole implementation in C!! for having better transactions):

runs like a cron job every 10 seconds..

That works well with Polaris Catalog - and Open Data Catalog - that syncs to/from pg_lage metadata:

# DuckLake on Hetzner

why Hetzner:

1/3 of AWS (as of now).
cost-sensitivity
and self-host

GitHub - berndsen-io/ducklake-hetzner: DuckLake on Hetzner for under 15 euros a month. · GitHub:

Also in Newsletter MotherDuck at DuckDB Ecosystem Newsletter – March 2026.

Downsides so far:

# DuckDB Powering Cancer Genomics Research

To give access to data, with DuckDB WASM:

This is how it works - using Keycloak and proxy:

# DuckDB as an Analytical Runtime (SQLRooms)

Building local-first analytics apps with SQLRooms

What is SQLRooms:

They have duckdb integration, and motherduck.

Pluggable and Composable (data stack?):

How to integrate:

They started to CLI tomake it more useful, with a DuckDB database.

This is how it looks:

~ Looks like a Notebook style, or sql
- ? similar to SQL Workbench Embedded from Tobi Müller?

# Laptop, a duck, and twenty years of wind

Auditing UK energy policy without a cluster.

Architecture in 2001 with Oracle and BusinessObjects:

Today in 2026:

Classical Medallion Architecture running on the laptop:

Writen the requirements in the “Kirkensen Syntax” - clean and self-documentation:

? What’s the syntax?

Looks something like this:

composed data pipelines:

The outcome is very detailed and can crunch much more data just for “breakfast”:

Nice take-away:

New way of working - Socio-Technical:

# Agenda

	Sponsor track (in-person)
14:15	So you self-hosted DuckDB. Now what? An honest look at the road to production	Mehdi Ouazza (MotherDuck)
	Main track, session 1 (in-person & streamed)
15:00	State of the Duck	Hannes Mühleisen and Mark Raasveldt (DuckLabs)
15:30	Effortlessly modernize your legacy: SQLFrame migrates PySpark to DuckDB without code changes	Nicolas Renkamp (Merck KGaA)
15:45	Phages and DuckDB: The mighty duo against the giants	Herminio Vazquez-Cano (ETH Zürich), Virginie Grosboillot (U. of Ljubljana)
16:00	ggsql: A grammar of graphics for SQL	Teun van den Brand (Posit PBC)
16:15	Coffee break
	Main track, session 2 (in-person & streamed)
16:35	How we built a SQL layer over our user listening history for agentic access	Kian Mehrabani (Spotify)
16:50	Grep your lakehouse: Search-first retrieval for DuckDB-powered agents	Sylvain Utard (Altertable.ai)
17:05	When the sea lion learns to quack: DuckDB as a MariaDB ystorage engine	Roman Nozdrin (MariaDB Corporation)
	Lightning talks
17:20	Native app development using DuckDB	Milla Henriksson (Toyota Gazoo Racing World Rally Team)
17:26	Making Iceberg walk and talk like Postgres	Marco Slot (Snowflake)
17:32	DuckLake on Hetzner: Your own data lakehouse for under €15 a month	Floyd Berndsen
17:38	DuckDB powering cancer genomics research	Thomas Visser (Hartwig Medical Foundation)
17:44	DuckDB as an analytical runtime: Building local-first analytics apps with SQLRooms	Ilya Boyandin (Foursquare / GeoVisually GmbH)
17:50	A laptop, a duck, and twenty years of wind: Auditing UK energy policy without a cluster	Barry Smart (endjin)
	Social event
18:00	Drink and snacks

Origin: DuckCon