Architecting Intelligence
Autonomous agents, language models on owned hardware, learned robotics. Tippel builds them as software for companies and develops its own systems in the lab.
Tippel/├── work/ AI software for companies│ ├── agents processes across systems│ ├── knowledge-systems answers with sources│ ├── model-adaptation open models, your data│ └── vision image, 3D, measurements├── lab/ our own systems│ ├── coding-agent 2839 passed│ ├── robotics Humanoid · MuJoCo│ ├── siem 192 passed│ └── tippelpi 204 passed├── notes/ 27 articles└── contact
For companies
- 01
Agentic systems
Multi-step agents with tool access to existing systems, schema-validated and traceable in the audit log.
- 02
Knowledge systems
Hybrid retrieval with source citations and a permission filter on every chunk.
- 03
Model adaptation
Fine-tuning open models on domain language, forms and standards.
- 04
Image and sensor data
Deep learning for image, 3D and measurement data, from classification to segmentation.
Our own systems
- Coding agent2839 tests passingAutonomous agent with a permission engine, an OS sandbox and undo for every change.
- Roboticslearned gaitHumanoid robot, designed in CAD, gait learned in MuJoCo.
- SIEM validation layer192 tests passingValidates KQL detection rules for Microsoft Sentinel before deployment.
- TippelPi204 tests passingLanguage model and hybrid retrieval on a Raspberry Pi 5, entirely inside the local network.
AI engineering from Hombressen
Tippel is an owner-managed AI engineering company. It was founded by Lukas Friedrich, a physicist specialising in deep learning. The standard for every project comes from research: a result only counts once it is measured and reproducible. Every project stays transferable from the first week, because source code, documentation and the deployment description live in the client’s repositories.
| Lead | Lukas Friedrich |
|---|---|
| Location | Hombressen near Kassel, Germany |
| Deployment | cloud API · EU cloud · your own servers |
| Contact | info@tippel.ai |