Artificial intelligence is changing the way we analyze, understand and use data – and we as the FIS Group are actively shaping this development. At what is now the fourth internal AI hackathon of the FIS Group, in which colleagues from FIS GmbH, FIS-ASP and Medienwerft GmbH took part, a big step was taken towards the future: the birth of FISMWASP-multimodal, our new next-generation AI model.
The hackathon took place in the offices of FIS GmbH in Grafenrheinfeld – an ideal location for creative ideas, focused collaboration and technological innovation.
Teamwork meets technology: how FISMWASP-multimodal was created
In order to master the complex requirements of a multimodal AI model, the participants relied on a clear division of tasks. The hackathon team split into two specialized groups:
Team 1 concentrated on data preparation. This primarily involved structuring a wide variety of data sources such as texts, images, sensor data or log files in such a way that they are suitable as input for an AI model.
Team 2 was dedicated to model development. The aim was to design and build FISMWASP V2 – an AI core model that processes and analyzes data across different modalities.
This structured approach made it possible to create essential foundations for the prototype in a very short time. The results speak for themselves – the first step towards a functional multimodal model has been taken.
What is FISMWASP-multimodal?
FISMWASP-multimodal – also known as FISMWASP V2 – is the evolutionary successor to our original FISMWASP, which focused exclusively on text data. The new version goes much further:
- Multimodal data processing: FISMWASP V2 can process different data types such as text, images, sensor data and log files simultaneously.
- Vector space mapping: All data is mapped in a common vector space – a prerequisite for intelligent, networked analysis.
- Similarity analyses: The model recognizes correlations and similarities across the boundaries of individual data types – a powerful tool for all data-based processes.
The advantages at a glance
FISMWASP-multimodal has a number of key features that make it a real innovation:
- Multimodality: The combination of different data sources enables new insights that would not be visible with individual data types.
- Real-time capability: analyses are provided quickly and efficiently – a clear advantage in time-critical applications.
- Conserving resources: The model was deliberately designed for efficient and high-performance processing.
- Innovation platform: FISMWASP V2 is not just a model, but the basis for numerous future AI features and products.
Our vision: New AI applications, new possibilities
What makes this model special is its ability to analyze similarities and correlations in real time across different types of data. This opens up completely new use cases in different areas:
Semantic search: Users find exactly what they are looking for – regardless of whether it is in text, images or other forms of data.
Personalized recommendations: By analyzing a wide range of user data, a comprehensive picture of interests is created – for highly relevant recommendations.
Anomaly detection in industrial processes: The combination of sensor and image data enables the early detection of deviations before failures occur.
What comes next?
The team not only developed a concept in the hackathon, but also initiated the first version of the prototype. The data processing strategy is in place, the architecture of the model has been defined – and the foundation stone for a future-oriented multimodal AI solution has been laid.
With FISMWASP-multimodal, we have a clear goal in mind: to connect data intelligently, enable new insights and create real added value – in real time and across all formats.