Data flows, operations and continuity.
This page answers the questions IT leadership and data protection officers put to an AI vendor: where data flows, what keeps running after a project and how handover is secured. Anything missing, we deliver on request: info@tippel.ai.
Data flows your DPO can sign off
For every deployment path you receive a data processing agreement (Art. 28 GDPR) and documented data flows. Your data is never used for training — on any path.
- 01
Cloud API
Your documents stay in your systems. Only the individual request travels, encrypted, to the provider: to OpenAI and Azure OpenAI via their EU endpoints, to Anthropic models via the EU regions of AWS Bedrock or Google Vertex AI, in each case under the provider’s DPA with its zero-data-retention options. Subprocessors are named in the DPA.
- 02
EU cloud, self-hosted
Models and data run in a German or EU data centre under our contract or yours. The only subprocessor is the hosting provider — named in the DPA.
- 03
Your own servers
Installed on your hardware, on request without any internet connection. There are no subprocessors; remote access only under your rules.
Technical and organisational measures (TOMs) and the DPA template are available on request. For financial institutions: the DORA datasheet covers the regulatory classification of the Sentinel validation layer.
What keeps running after the project
The concrete numbers for your case come out of the Readiness Check — the structure is always the same:
| Cost type | Cloud API | EU cloud | Own servers |
|---|---|---|---|
| One-off | — | Setup | Hardware (GPU servers) + setup |
| Running, usage-based | API fees per processed request | — | — |
| Running, fixed | — | Monthly server costs | Power, rack space, depreciation |
Continuity: what happens if the engineer is unavailable
Tippel is an owner-managed one-person company — you work with the engineer who writes the code. Tippel answers the availability risk with handover capability, not reassurance:
- 01
Handover-ready from week one
From the start, you hold working, documented software in your own repositories. It rests on widely used tools, so another engineer can take over.
- 02
Standard stack, your infrastructure
Open, widely used tools instead of homegrown frameworks. Deployment is described reproducibly, credentials live with you — as do source code and rights.
Master data and documents
| Company | Tippel — sole proprietorship, owner Lukas Friedrich |
| Address | Kampweg 4, 34369 Hofgeismar, Germany |
| Tax number | 026 819 040 02 |
| Commercial register | Not registered (sole proprietorship) |
| Contact | Lukas Friedrich · info@tippel.ai · +49 178 5879849 |
| References | BLUEROPE Consult GmbH (client project) — details |
- DPA templateData processing agreement per Art. 28 GDPRon request
- TOM overviewTechnical and organisational measureson request
- DORA datasheetRegulatory classification of the Sentinel validation layeronline