48.85°N · 2.35°E — Sovereign data platform

Polnor

The European Lakehouse, on your cloud.

Compute, SQL, notebooks and ML on one console — your data, your keys, your jurisdiction. A complete data platform that answers to European law.

/Compute
Search resources⌘K ◖ Cardinal AM

Compute

Instances running on your cloud

+ New compute
Active compute
6
running instances
GPU hours · MTD
1,284
+18% vs last month
Spend · MTD
€ 9,420
across 3 providers
Running jobs
3
live · refreshing 30s
NameProviderSpecsStatusRegionAuto-stop
training-a100fine-tuning · team-ml OVH 8× A100 · 96 vCPU running GRA9 ◷ 2h
etl-spark-04nightly pipeline OVH 32 vCPU · 256 GB running GRA9 ◷ 30m
notebook-edaad-hoc analysis SCW 8 vCPU · 32 GB starting SBG5 ◷ 1h
warehouse-biduckdb + spark OVH 16 vCPU · 64 GB stopped GRA9 ◷ never

SQL Editor

DuckDB reads · Spark writes · warehouse: analytics

1
2
3
SELECT slug, name, count(*) AS rows
FROM acme.platform.workspaces
GROUP BY 1, 2 ORDER BY rows DESC;
Press Run to execute on your warehouse

Jobs

Multi-task DAGs · last 24h

+ New job
daily-pipelineran 12:04
ingest transform report
feature-backfillran 08:30
extract join write
churn-retrainran 06:00
prep train eval register

Experiments

MLflow-compatible tracking

Compare runs
ExperimentBest runMetricProgress
churn-v3 run_8f2a auc 0.934 step 4,200
fraud-lgbm run_1c7d auc 0.971 step 9,800
ltv-reg run_44b0 rmse 12.4 step 2,100

Notebooks

Git-backed · branch main

+ New notebook
[1]
import polnor as pl
df = pl.read("acme.sales.orders")
[2]
df.groupby("customer").amount.sum()
customer amount
acme 42,019.50
globex 18,402.10
[3]
from polnor import mlflow
mlflow.log_metric("auc", 0.934)
control plane — OVHcloud GRA9, France data plane — your VPC · your S3 · your VMs lat 45.7578° · lon 4.8320°

§ 01 — Capabilities

One chart for the
whole data stack.

01 ◇

Iceberg Lakehouse

Open Apache Iceberg tables on your own object storage. ACID, time-travel, schema evolution — no proprietary format, no lock-in.

Lakekeeper REST catalog
02 ◇

SQL warehouses

DuckDB for fast reads, Spark for writes and maintenance. Warehouses spin up on demand on your cloud and stop when idle.

DuckDB · Spark · on-demand VMs
03 ◇

Notebooks & jobs

Git-backed notebooks and multi-task job DAGs with dependencies, retries and per-task timeouts. Logs streamed live.

DAG · retries · live logs
04 ◇

ML & model serving

MLflow-compatible tracking, experiments, model registry and one-click serving endpoints — drop-in Python SDK.

polnor.mlflow · endpoints
05 ◇

Governance & lineage

Per-workspace isolation, audit log with S3 export, column-level lineage captured from every query and run.

audit · lineage · quotas
06 ◇

Bring your own cloud

Your OVH or Scaleway account, your VPC, your bucket, your security groups. We orchestrate; your data never leaves home.

OVH · Scaleway

§ 02 — Digital sovereignty

Tout est chez vous.
Everything stays home.

An enterprise-grade data platform that answers to European law — because the data, the keys and the metal are all on this side of the Atlantic.

000°

Data never leaves Europe

Compute and storage run in your own European region — OVHcloud or Scaleway. No transatlantic hop, ever.

060°

Immune to the US Cloud Act

No US-owned entity in the path. FISA 702 and the Cloud Act cannot compel access to infrastructure that isn't theirs.

120°

GDPR by design

Per-workspace isolation, full audit trail, data-residency you can point to on a map. Built for DPOs, not against them.

180°

Your keys, your bucket, your VPC

Credentials encrypted at rest. We orchestrate the control plane; the data plane is entirely yours.

240°

Open formats, no lock-in

Apache Iceberg and Parquet on object storage. Walk away with your tables intact, any day.

300°

Self-hosted everything

Even this website's fonts are self-hosted — not one byte calls a US CDN. Sovereignty is a habit, not a checkbox.

§ 03 — Architecture

Two territories,
one bearing.

Control plane

Polnor — OVHcloud GRA9 API · scheduler · catalog · audit · billing. Hosted in France, by us.
  • orchestration
  • governance
  • metering

Data plane

Your cloud — your region Agent · DuckDB & Spark sidecars · your VPC, bucket and VMs. Data never leaves.
  • your S3
  • your compute
  • your keys
  1. 01

    Connect

    Bring your OVH or Scaleway credentials. We provision a namespace and an agent in your cloud.

  2. 02

    Catalog

    Point Polnor at your S3 bucket. Iceberg tables register in the Lakekeeper catalog — your data stays put.

  3. 03

    Compute

    Spin up warehouses, notebooks and jobs on demand. They run on your VMs and stop when idle.

  4. 04

    Operate

    Audit, lineage, quotas and billing — one console, full visibility, European law end to end.

§ 04 — Pricing

Pay only for what
you actually use.

To start €0

No subscription, no credit card. Connect your cloud and go.

You pay your cloud

OVH or Scaleway bill you directly for the compute and storage you use.

We add a transparent margin

A clear multiplier on top — itemised against your provider invoice, no surprises.

Billed manually, in euros, by a European entity — invoices you can hand straight to finance.

Set a course

Request a
demo.

See your own Iceberg tables, warehouses and notebooks running on your European cloud — usually within a week.

hello@polnor.net · OVHcloud GRA9, France 🇫🇷

Opens your mail client — no data leaves your browser until you hit send.