A single ship listing dramatically against breaking storm clouds and white-foam waves.

vasa

The European data and AI platform.

Ingest, transform, analyse, infer.
All in one place, all inside Europe.

Start free

What you can do

One open lakehouse. Four jobs.

Iceberg storage, Trino analytics, native LLM inference, agent built in. All on bare-metal European compute. Priced like one platform, not five vendors stitched together.

Ingest

Push events to our streaming ingest API, schema-validated and deduplicated. Or pull from PostgreSQL, MySQL, MariaDB, SQL Server, Oracle, ClickHouse, SingleStore, Redshift, or MongoDB on a schedule.

Transform

Trino-powered SQL on Iceberg tables. Save SQL as transforms that materialise into tables. Schedule and chain them. Lineage and dependencies tracked automatically.

Analyse

Connect from your notebooks, dashboards, and BI tools. Same data, no copies, no replicas.

Inference & RAG

Call AI directly from your SQL. Generate embeddings on any column. Run RAG over your own data. All on EU-hosted infrastructure.

The agent

The agent can do anything you can.

Describe what you want. It lists your connections, builds the transform, sets the schedule, and asks before running.

vasa agent panel: a user prompt asks the agent to create a rollup table on GitHub data and add it to a scheduler. The agent's response confirms the rollup transform was created with explicit field aggregations and metrics, plus a daily schedule.

AI in SQL

AI lives where your data is.

Generate, classify, extract, embed, retrieve. From any SQL query, on GPUs we operate in Europe.

vasa query editor: an SQL query calls llm.ai.ai_gen() on each row of a GitHub events table to generate a one-line description of what each organisation builds. The results panel shows the org, the event count, and the LLM-generated description, returning in seconds.

Orchestration

SQL transforms with lineage and schedules.

Save SQL as transforms. Lineage tracked automatically. Schedule on cron: bronze, silver, gold.

vasa Jobs page: a horizontal lineage diagram shows a bronze_github_events transform feeding into silver_org_activity, which in turn feeds into gold_org_summary. Below the lineage, the transforms list shows descriptions, statuses, and last-updated timestamps. Below that, three daily schedules are shown with their cron expressions and Enabled toggles.

Who built it

Built by people who have done this before.

The team has engineered data platforms used by hundreds of millions of users globally, at Spotify and Kivra. We've shipped data applications under GDPR, financial-services compliance, and healthcare data requirements.

We know what enterprise scale looks like, what regulators ask for, and where the platforms you've used before fall short.

Sovereignty

Your data never leaves Europe.

This is the one thing that separates us, and it's not marketing. We never touch American companies or American infrastructure. Here is the structure, layer by layer.

Storage
Hetzner Cloud, in Germany and Finland. Iceberg tables, object storage, Kafka brokers, the catalog, the activity log: all of it.
Inference
Open-weight models on GPUs we operate ourselves, in Europe. No US-hosted model API in the path. No third-party "we send your prompts to a partner" middleware.
Vectors
A managed Qdrant cluster on the same EU infrastructure as your tables.
The agent
Runs in Europe. The audit trail is in Europe. Your queries, your prompts, your embeddings, your logs: all in Europe.
Aerial view of a snowy northern landscape, with distant mountains and golden-hour light across a forest of frost-covered trees.

Pricing

Usage-based. No seats. No surprises.

Start free. Simple, usage-based pricing after that. You pay for what you use, nothing more. No per-seat licensing. No annual commits. No surprise quarter-end bills.

FREE

Free tier

€0 forever

  • Generous-enough storage and compute to actually evaluate
  • Self-serve sign-up, no card required
  • Public docs, community Discord

PAY AS YOU GO

Usage-based

Per query · per GB · per token

  • Trino compute, billed per query-second
  • Iceberg storage on Hetzner, per GB-month
  • LLM inference, per million tokens
  • No minimums, no annual commit

Two ways to start

Sign up tonight. Or talk to us first.

Start free.

Sign up, ingest data, run your first query in under an hour. No card, no sales call, no procurement form.

Start free →

Talk to us.

Walk through the platform, talk through your sovereignty constraints, see if vasa is a fit. Direct line to the founders. No SDR layer.

Talk to us