1998  

During my studies I worked several years for the German Research Center for Artificial Intelligence (DFKI GmbH). As a scientific assistant I worked together with Harald Meyer auf'm Hofe in the ConPlan-project. We created a solver for Hierarchical Constraint Satisfaction Problems (HCSPs). The solver was written completely in C++ and runs under Windows (NT) and Unix / Linux. The solver is able to do local search, too, which is essential for coping with extremly large search domains.
In addition to my job as student assistant, I worked for several other enterprises where I was given the authority to develop software components in C++ and Delphi.

1998  

In form of a student project (Projektarbeit) I realized a Speech Recognition of Singular Words for Controlling a Robot (1999) [Acrobat PDF (german)]. I have always been interested in utilizing artificial intelligence to solve real-life problems. So, in the speech recognition project, for example, I applied case-based reasoning and a time-warping similarity measure to recognize a spoken word by comparing it to previously stored speech samples.

2000  

I studied computer science and mathematics at Kaiserslautern University (Germany) and received the Diploma (master's degree) in August 2000. In my master thesis (August 2000) I investigated Weakly Structured Workflows for the Knowledge Management in Enterprises and implemented a tool for executing and monitoring these weakly structured workflows. This was an introductory work for the weakly structured workflows paradigm which I extended later as a full researcher and member of the FRODO project.

2000  

I entered the DFKI as a researcher.
I started my professional research carrer in the FRODO project where I realized the weakly structured workflow paradigm. The workflow management system (WfMS) was realized in form of a multi-agent system. New agents fulfilling new tasks could be integrated easily into the system, e.g., information agents can be plugged in that way.
The information processed by the agents were formed using multiple ontologies which enabled sophisticated inferencing.
We started implementing a graphical user interface (GUI) called FRODO TaskMan. Although this GUI was a powerful research tool a short user study showed that it is too overloaded for a normal knowledge worker. The problem is, that the GUI provided direct access to all workflow modeling aspects. I worked out possibilities to reduce some of the main cumbersome interactions with the WfMS. One of them lead to the research in the follow-up project EPOS:

2003  

In the EPOS project I investigated the automatic elicitation of a knowledge worker's context using user observation. One of the goals was to utilize the estimated user context to identify potentially relevant workflow tasks. That way we can realize an implicit workflow management interface: If the system can guess which workflow task is currently processed by the user (and the user aknowledges that), the main workflow interaction is just a question of one or two mouse clicks. Instead of manually viewing the whole work item pool (workflow inbox) and looking which tasks have to be processed, the system proposes relevant workflow tasks matching his current context.
Another important - and actually the most important - goal for estimating the knowledge worker's context is the realization of context-aware assistance. We enriched a simple note taking tool with a context annotation facility. Notes taken down by the user are annotated with the user's current context. This creation context can be used for context-sensitive retrieval, filtering, or pro-active delivery. A multi-criterial, pro-active, context-sensitive information delivery has been realized in the EPOS Assistant Bar which shows documents, topics, organizational entities, and workflow tasks which are estimated as relevant for the user's current context.