Skip to content

Sovereign data & AI engineering

Run AI on your most sensitive data. Without giving up control.

Datasider designs, builds and operates sovereign data and AI platforms for sensitive data — on your infrastructure or a European cloud. Open source, yours to keep.

Data platforms · AI agents · Trusted research environments

30 minutes · With the architect who will lead your project · No sales pitch

Haven · Research workspace

Isolated

Output review · before any export

Agent

Knowledge
assistant

Answers from
your documents

Data stayed in

Sovereignty status

All controls active
  • Data location

    France

  • AI model

    Mistral · running on your servers

  • External AI API calls

    0

  • Encryption keys

    Held by you

  • Exit strategy

    Documented, as code

Is this you?

Your data can’t just leave.

Datasider is built for organisations facing one or more of these constraints.

01

AI is blocked by security

Sensitive data cannot be sent to public AI providers.

02

Your data stack is fragmented

Data lives across systems, teams and providers, with no unified governance.

03

AI pilots never reach production

The proof of concept works. Security, infrastructure and compliance stop the rollout.

04

You need sovereign control

Jurisdiction, hosting and operational access matter as much as performance.

05

You are becoming dependent on a vendor

You need open standards, portable workloads and an exit strategy.

See how we solve this

Solutions

Three ways to put your data to work.

Core is the foundation. Agents put AI to work on it. Haven brings AI to sensitive data that can’t move.

01 / Foundation

Datasider Core

Your data foundation.

A sovereign, open-source data platform on your infrastructure or a European cloud. We build it, run it, or hand it over.

ingestionlakehousegovernancesecurity

e.g. health data warehouses, research data platforms, governed lakehouses.

02 / AI

Datasider Agents

Put AI to work on your data.

AI agents built on your governed data, with open-weight models, controlled access and every action logged.

Examples of what we build

  • knowledge assistants
  • automated reporting
  • anomaly detection
  • request triage

An agent is only as good as the data it can reach. We build both.

Design partner programme

03 / Sensitive data + AI

Datasider Haven

Your data stays. AI comes to it.

A Trusted Research Environment (TRE) to analyse, collaborate and run AI on sensitive data — without moving it.

Discover Haven

How we work

From strategy to production.

We help you frame the right use cases, organisation and roadmap — then we build the platform, put it in production and stay accountable for it, until your team is ready to take over.

  1. 01

    Frame

    Use cases, organisation, governance, roles and roadmap.

  2. 02

    Architecture

    Designed around your data, constraints and infrastructure.

  3. 03

    Build

    Implemented on open-source components, as code.

  4. 04

    Deploy

    On your servers, on a sovereign cloud, or air-gapped.

  5. 05

    Operate

    Run in production by the engineers who built it.

  6. 06

    Handover

    Documented and handed to your team when you’re ready.

You work directly with the architect who designs your platform.

Depending on the project, Datasider brings in a network of senior data, DevOps and AI engineers — coordinated and accountable under one contract.

Sovereignty

Sovereign by design.

Your data stays on your infrastructure or in Europe. You choose the jurisdiction it falls under.

Sovereign cloud.

Deployed on the European provider of your choice — including SecNumCloud-qualified and HDS-certified providers — or on your own servers.

Open source.

Proven open components you can inspect, run and extend.

No vendor lock-in.

Open standards. Take it all back whenever you want.

Reversibility, built in

  • →Your data remains in open formats.
  • →Your workloads run on standard infrastructure.
  • →Your models can be replaced.
  • →Your architecture does not depend on Datasider.

You own the architecture. We help you build it.

Sovereignty in practice — 10 concrete controls
01
Infrastructure ownership
The platform runs on servers you own or rent. Not on ours.
02
European hosting
When you use cloud, it is a European provider of your choice.
03
Jurisdiction
You choose the provider, and with it the legal framework your data falls under.
04
Open source
Every component can be inspected, patched and run without us.
05
Portability
Kubernetes and open formats: move between on-prem and cloud without a rewrite.
06
Access control
Identity runs on your own identity provider. You decide who sees what.
07
Encryption
Data encrypted in transit and at rest, with keys and secrets managed on your side.
08
Operational access
Our engineers access your systems only as you allow, and that access can be withdrawn.
09
Vendor independence
No proprietary layer in the middle. No licence that holds your data hostage.
10
Exit strategy
Documented, as code, handed over. Leaving us is a planned step, not a project.

Exact controls depend on your deployment and are agreed with your security team.

Read our sovereignty & security approach →

Proof

Selected experience

Core

Hospital health data warehouse

Context
French university hospital
Challenge
Make sensitive health data available for analytics and research while keeping it under the organisation’s control.
What we built
A sovereign data platform on infrastructure the organisation controls: lakehouse, governance, security and AI readiness.
Result
Running in production on sensitive health data.
Core

Scoping a national health data warehouse

Context
National health data consortium
Challenge
Define the functional and technical architecture of a specialised health data warehouse, as the basis for a public funding application.
What we do
A series of scoping workshops leading to a single functional and technical architecture document.
  1. 01 Store

    S3-compatible object storage · Apache Iceberg · PostgreSQL

  2. 02 Process

    Apache Spark · Trino · DuckDB

  3. 03 Transform

    dbt · SQLMesh

  4. 04 Orchestrate

    Apache Airflow · Dagster

  5. 05 Govern

    Apache Polaris · OpenMetadata

  6. 06 Secure

    Keycloak · OpenBao

  7. 07 Analyse

    JupyterHub · RStudio · Apache Superset

  8. 08 AI

    vLLM · open-weight models (Mistral, Llama, Qwen) · pgvector · Qdrant

  9. 09 Run

    Kubernetes · Argo CD · OpenTofu — on-prem, sovereign cloud or air-gapped

Who you’ll work with

Portrait of Quentin Richard, founder of Datasider

Quentin Richard

Founder & Lead Architect

Quentin spent 10+ years building data platforms in production — as data engineer, data architect and data project director — before founding Datasider.

LinkedIn ↗

You don’t get an account manager. You get the engineer who designs the system.

FAQ

Questions we often get.

Yes. We deploy on the HDS-certified provider of your choice, or on your own infrastructure.

Your data stays in open formats, your workloads run on standard infrastructure, and everything is documented as code. Leaving is a planned step, not a project.

Not by default. Open-weight models run inside your environment. External APIs are only used if you explicitly choose them.

Yes: your Kubernetes clusters, your existing data stack and your identity provider.

Yes. Everything is documented, as code, and handed over when your team is ready.

Contact

Keep control. Ship to production.

Talk to an engineer. 30 minutes, your context, your constraints. No sales deck.