Skip to main content
Private EU GPU · Healthtech & sensitive data

Run your sensitive ML workloads on EU soil. Zero DevOps.

GhostNexus is a Python GPU execution API hosted in Frankfurt, built for healthtech and regulated teams that need to process ML workloads inside the European Union — with a GDPR Art.28 DPA available within 24 h.

For: CTOs · DPOs · Technical founders · Heads of Data / AI

Frankfurt
Hetzner datacenter, DE
24 h
GDPR Art.28 DPA
€299
Pilot / month · 5 seats
📍EU-hosted · Hetzner Frankfurt🇫🇷French registered business — SIRET 102 801 883🔒GDPR Art.28 DPA available🗑️Data deleted after each job

Your ML pipeline may have a GDPR blind spot.

You use an international GPU platform to run your models. It works. The problem: international GPU platforms can involve region, sub-processing or transfer choices that are hard to document for a regulated team.

Data processing agreement

The GPU platform does not provide a GDPR Art.28 DPA — your DPO blocks the deployment.

Data location

Impossible to confirm in writing that data never left the EU during execution.

In-house GPU infra

Building compliant GPU infrastructure internally: 3 to 6 months. Meanwhile, the model waits.

GhostNexus reduces this to 24 hours. You submit your use case, you receive the DPA and pilot access. Your workloads run in the EU without you managing any infrastructure.

Managed GPU execution. Data in Europe. GDPR documentation included.

We host the GPUs. We manage the infrastructure. We provide the Art.28 DPA.

ghostnexus — job submit
import requests

response = requests.post(
    "https://api.ghostnexus.net/api/jobs",
    headers={"x-api-key": "YOUR_KEY"},
    json={
        "task_name": "inference-pipeline",
        "script_content": open("my_model.py").read(),
    }
)
# → Runs on a Frankfurt GPU · Data deleted after the job
  1. 1
    Account + API key
  2. 2
    Submit script + data
  3. 3
    GPU execution in Frankfurt
  4. 4
    Retrieve results
  5. 5
    Data deleted

What you don't have to manage

  • No server to provision
  • No CUDA, Docker or Kubernetes setup
  • No data processing agreement to draft from scratch
  • No hosting audit to initiate

What you get immediately

  • GDPR Art.28 DPA signed by GhostNexus
  • Written confirmation of processing location (Frankfurt, DE)
  • Documented REST API
  • Concierge onboarding for early customers
  • Direct support — response within 24 business hours
  • DPA and pilot access within 24 h after use-case validation

The hardware, precisely.

You can't evaluate a GPU offer without the specs. Here they are.

Pilot — dedicated GPU
NVIDIA RTX 4000 Ada
  • · 20 GB GDDR6 ECC VRAM
  • · Dedicated server — never shared between clients
  • · Hetzner datacenter, Frankfurt, Germany
  • · CUDA 12, PyTorch / HuggingFace preinstalled
What 20 GB covers in practice
  • · LoRA / QLoRA fine-tuning of 7-8B models (LLaMA 3, Mistral)
  • · Inference of models up to ~13B quantized
  • · Embeddings, classification, medical vision (CNN, ViT)
  • · Batch pipelines: scoring, clinical NLP, image processing

Need more (70B, multi-GPU)? Extended capacity on quote — same Frankfurt location, same DPA.

We don't publish theoretical benchmarks copied from vendor datasheets. Measured execution times on your real use case are provided during onboarding, before any commitment beyond the first month.

What GhostNexus is not.

Clarity avoids bad surprises — especially for your DPO.

What it is ✓
What it is not ✕
Python ML execution API on GPU
A permanent health-data hosting provider
Workloads processed on EU infrastructure
An HDS certification — we do not hold it
GDPR Art.28 DPA available immediately
A guarantee of absolute compliance for your legal context
Designed to reduce non-EU processing risk
A substitute for your DPIA or your legal counsel
GPU alternative with documented legal framework
A raw rented GPU server with no documentation

Note: if your use case involves long-term storage of personal health data or strictly requires HDS certification, we will tell you before starting. Some cases do not fit our offer.

Paid pilot. Five seats. No commitment.

No mandatory sales call. DPA and access provided within 24 h after use-case validation.

FOUNDING RATE · 5 SEATS
€299/month

No commitment · Cancel at day 30

  • Full API access
  • Hetzner Frankfurt GPU
  • Signed GDPR Art.28 DPA
  • Concierge onboarding
  • Direct support — response within 24 h
  • Usage volume defined at onboarding
Request this pilot →
Standard rate
€499/month

From the 6th customer

  • Everything in the pilot
  • Extended volume
  • Negotiable SLA
  • Monthly invoicing
Contact us

What we state — and what we don't promise.

We are direct about this because your DPO will be too.

Processing location

Workloads submitted via the API run exclusively on Hetzner servers, Frankfurt datacenter, Germany — European Union.

GDPR Art.28 DPA

Provided within 24 h, signed by GhostNexus, based on GDPR requirements and informed by CNIL-recommended clauses.

Data retention

Data submitted for execution is deleted after each job. Logs are kept 30 days and deletable on written request.

Sub-processors in the execution chain

No non-EU sub-processor is involved in executing your ML workloads. Billing, email and support services are documented separately.

Data access

No human access to workload data by default, unless explicitly requested by the client for support or debugging.

What we do not claim

  • We are not HDS certified. If your case strictly requires it, we will tell you before any commitment.
  • We do not replace your DPIA. We can provide the technical elements you need to complete it.
  • We do not guarantee absolute compliance of your processing — that depends on your overall architecture.

GhostNexus is designed to help regulated teams reduce exposure to risks linked to processing sensitive data outside the European Union. This text does not constitute legal advice.

“What if you disappear?” — fair question.

GhostNexus is an independent French business. You have every right to ask what happens if it stops. Here are the commitments, in writing.

Zero technical lock-in

Your scripts are standard Python (PyTorch, HuggingFace). No mandatory proprietary SDK. Migrating to OVH or in-house infra = changing one API URL.

Guaranteed reversibility

Export your data and results at any time, free of charge. If the service ever shuts down: 60-day written notice and migration assistance, in the contract.

Nothing to recover

Workload data is deleted after each job. In practice there is no data stockpile at GhostNexus you could be held hostage over.

Live service status: ghostnexus.net/status — monitored automatically every 30 minutes.

Frequently asked questions

Do workloads stay in Europe?
Data and scripts submitted via the API are processed on Hetzner infrastructure in Frankfurt, Germany. No non-EU sub-processor is involved in execution. Peripheral services (billing, email) are documented separately.
Are you HDS certified?
No. If your use case strictly requires HDS-certified hosting, we will tell you clearly before any commitment. If your usage covers ML workloads with pseudonymized or anonymized data, the applicable requirements may differ — clarify this with your DPO.
How does the DPA work?
We send you a GDPR Art.28 DPA within 24 h. It covers: nature and purpose of processing, security measures, processing location, retention, and data-subject rights. Your legal counsel or DPO can review it before signature.
What happens after the pilot?
The pilot has no commitment. After the first month you continue at €499/month or you stop. No automatic renewal without explicit agreement.
Can we use our own models?
Yes. You can use your own models and libraries compatible with the provided execution environment (PyTorch, HuggingFace, scikit-learn, etc.).
Why a flat monthly rate instead of usage-based billing?
A flat rate simplifies your accounting and avoids surprises. The pilot includes a usage volume defined together at onboarding. For exceptional volume, a dedicated quote is proposed.

Request a pilot — €299/month

5 seats at the founding rate. DPA and access provided within 24 h after use-case validation. No commitment.

Get the DPA first

DPA and pilot access provided within 24 h after use-case validation. 5 seats at the founding rate (€299/month).