Process-Automation with AI support.
Your Data, Your AI: Secure, private und powerful.
thalabus: connecting human and artificial skills.
We are building AI solutions for our customers - fast, secure and elegant with the LLMs that are best for the purpose.
- Our Chatbots can access internal and external systems
- Solutions for semantic search, knowledge management and data quality
- Workflow-automation with AI support
"Standard" Chatbots for websites and customer support teams are easy to create. When requirements get more complicated, a more robust, powerful and flexible solution is needed
We are integrating systems, data and knowledge since 25 years. No wonder that our thalbus can handle documents but also information from all kind of sources, such as customer backends, databases, Intranet, Internet. We create custom-tailored ontologies and knowledge-graphs from there. Our customers' AI solutions can understand and solve complex questions & problems.
thalabus can spot inconsistencies and contradiction in your documents and other data sources, and automatically suggest corrections: an invaluable tool to increase the quality of your data.
AI solutions based on thalabus raise customer support and employee productivty on a completely new level.
For data privacy-reasons thalabus was designed from the very beginning to produce good results with small LLMs. Those can run at low cost in your local data center or your private cloud. Your data stays safe in Switzerland that way.
Your t2b consultant will analyse the requirements and business processes with your team. He/she will discuss enhancements and changes with the team and then communicate the desired solution to your thalabus instance.
t2b will make sure that the efficiency gains made possible with AI support are effectively realized.
How can thalabus do all that?
thalaplan: this is the core element of the system. It breaks down intelligently complex tasks in multiple, smaller steps. These can then be solved also by agents using smaller size LLMs.
complex RAG: when thalabus is ingesting documents and information it not only tags them, but stores meaning and relations in ontologies and graph-DBs. With this 1.) the system can verify that the data is correct and 2.) the user can validated any answer with the direct link to the underlying sources.
99+% correct answers: all results from intermediate steps and also generated code is tested automatically on completeness and correctness. With this we can exclude erroneous answers. As we deliberatly choose to not use external code and frameworks, the thalabus code can be tested at 100%..
Natively multi-lingual: we are using native D/F/I/E LLMs for the user/language interaction and only translate the "technical" work to English. Other languages are available on request
Extensibility: thalabus components have several extension points. Those can be used by thalabus itself, t2b consultants and customer teams to extend the functionality as desired.
thalabus is getting configured and parametrized via natural language. If required, t2b will also fine-tune small LLMs to specific tasks/specialisations.
thalabus-solutions: elegant, secure, correct, and efficient.
