Services
From the analysis to the running system.
Five areas we offer individually or as one continuous project. Scope and pace follow your goal.
- 01
Consulting & technology advisory
Before we build, we clarify what should be built.
We assess ideas, existing systems and data and deliver a sound estimate of effort, risk and value. That includes the honest answer when a project is not worth doing.
- Feasibility and opportunity assessment
- Technology choice: cloud, hybrid or local
- Architecture, data flow and interfaces
- Cost, operations and scaling models
- Business case and roadmap
- Workshops
- Architecture design
- Requirements
- Prototype
- 02
Software development
We build applications that last.
From prototype to production: we develop web applications, backends and interfaces, integrate them into your landscape and document them so they stay maintainable.
- Web frontends and backends
- APIs and databases
- Mobile and desktop applications
- Integration into existing systems
- Test coverage, error handling, documentation
- Python
- TypeScript
- Node.js
- FastAPI
- PostgreSQL
- Docker
- 03
AI and large language models
Language models built in, not bolted on.
We integrate large language models into your products and processes – with clear responsibilities, measurable quality and controllable cost.
- LLM integration into existing applications
- Retrieval-augmented generation (RAG) over your data
- Prompt, testing and evaluation strategy
- Cost control: caching, routing, smaller models
- Quality assurance and monitoring
- OpenAI-compatible APIs
- RAG
- Embeddings
- Vector databases
- Evaluation
- 04
Local models & data protection
Run models where the data lives.
When sensitive or personal data is processed, we run models on your own hardware or in your private cloud. Your data does not leave the building.
- Operation with Ollama, llama.cpp or vLLM
- On-premises and private cloud
- No transfer of your data to third parties
- GDPR-compliant processing
- Setting up, selecting and tuning hardware
- Ollama
- llama.cpp
- vLLM
- Linux
- NVIDIA
- Docker
- 05
Agents & automation
Agents that do more than answer.
We build systems that research, extract information, verify results and trigger process steps on their own – with the guardrails a production system needs.
- Agents with tools (tool use)
- Multi-step workflows and approvals
- Automation of recurring processes
- Guardrails, logging and control
- Connecting to existing systems via APIs
- Tool use
- Workflows
- APIs
- Queueing
- Logging
BERLITEC
Which area fits your project?
Describe your goal in a few lines. We will tell you the shortest route.