Generative Artificial Intelligence
With the availability of huge amounts of data, automated and intuitive ways to query and visualize this data and design dedicated workflows are required to efficiently use the rich potential of the collected information. Today, generative artificial intelligence (GenAI) algorithms like large language models are the state-of-the-art tools to reach these goals.
Data only generates value combined with efficient ways to extract information and guide decisions. The two important paradigms in this context are intuition and automation, both drivers for the success of GenAI in recent years: Large language models allow users to intuitively search for information in their natural language instead of using programming or query language syntax; and they allow the automation from small tasks to whole workflows.
We research which types of large language models perform best for our use cases, with what data and how to train them, and how to optimally execute fine-tuning and instruction tuning. Other topics are methods of ingesting external information (internal data, documentation or chats), prompt engineering, and the performance evaluation of systems including large language models.
Some of the current use cases for the application of our research results include the automated creation of queries from natural language; the explanation of complex queries; support and onboarding to Dynatrace services via chatbots; and the automated generation of data visualization and dashboards as well as of entire workflows.