AI search allows users to retrieve both general and company-specific information in natural language. The difference with traditional search methods: You don't have to formulate your query exactly because AI recognizes the intent of the query and delivers the most relevant content. And you can also add questions to your query to receive further details or explanations.

For AI search, we use general language models such as ChatGPT or open-source models such as Llama,. We combine these with Retrieval Augmented Generation (RAG) that allows semantic information from texts to be stored in a vector database. To answer user queries, the language model prompt is automatically enriched with suitable content from the database.

Our AI solutions specialize in efficiently extracting data from a wide variety of documents. A common application scenario is the listing of requirements from specifications that may be in different formats, such as a PDF document or handwritten. Our AI solutions are able to capture this data and convert it into a consistent, structured presentation, such as an Excel spreadsheet. We rely on our customized data pipelines, which greatly simplify the process of data extraction and transformation.

Depending on the requirements, we use different technical methods. For data extraction from unstructured sources, we typically use language models. For pure data transformation, such as the conversion of one measurement data format to another, we recommend efficient conversion algorithms. These can be implemented using a language model optimized for code generation, such as Python.

If you don't want to entrust your sensitive or personal information to the Internet, you still have smart help. Are you looking for the right wording for an email? Does a Word document need to be translated into English? Do you need to summarize a large document or create a formula in Excel? Our AI assistants simplify and streamline complex office tasks. These assistants are integrated directly into the MS Office tools Word, Outlook and Excel, allowing you to customize content with just a few clicks.

Our Office plug-ins, with their built-in features, use powerful AI models internally that are optimized for the task at hand. For example, to improve the writing style in documents, the desired writing style is compiled with the text as a prompt for a language model, and the plug-in then applies the result directly to the Office document.

Our AI-driven coding provides comprehensive support for all phases of software development - from device control programming to complex enterprise data analysis. By integrating coding plug-ins for popular development environments such as IntelliJ IDEA, Visual Studio, or Eclipse, we make it easy to generate code, unit tests, documentation, and database scripts. In addition, our tools allow you to perform code reviews to find bugs or security vulnerabilities. And our AI models can learn to use your proprietary libraries and frameworks as well as open source libraries.

For AI-based coding, we use open-source plugins for IDEs, which are often connected to an on-premise hosted language model. As a model, we use language models optimized specifically for code generation, such as the WizardCoder. With the help of Retrieval Augmented Generation (RAG) and fine-tuning, we extend the AI model with company-specific knowledge. For fine-tuning, we use a self-developed pipeline that also takes care of the generation of test data.