Transfer & public funding
We develop applied AI systems in publicly funded and industry-oriented transfer projects. The focus is on robust demonstrators, understandable AI architectures, and solutions that must work in real processes across construction, healthcare, circular economy, retail, and IoT security.
AI copilot for sustainable architectural planning
AI-BIM develops an AI-supported copilot that makes sustainable architectural designs plannable and optimizable while taking environmental standards into account.
Our role is to develop and integrate the copilot architecture so that AI models support concrete planning decisions instead of only generating design variants.
Multimodal BIM copilot for machine-readable instructions
Show2Instruct translates natural-language interaction and visual object references into machine-processable control commands for BIM contexts.
We integrate foundation-model-based AI architectures and work on reliable multimodal BIM copilots that connect domain knowledge, model data, and tool calls.
AI prototypes for emergency dispatch centers
LeLaMar investigates AI use cases for emergency dispatch centers to structure emergency-call data faster, recognize incident situations, and reveal links between multiple calls.
We develop and evaluate proof-of-concepts for transcription, risk analysis, event linking, event tracking, and ad-hoc incident summaries.
Autonomous on-demand transport systems for hospitals
MediCar 4.0 develops highly automated electric vehicle systems and modular containers for safe on-demand transport processes in hospitals.
Our contribution is robust multi-agent coordination of autonomous vehicle fleets and safeguarding their behavior in dynamic hospital corridors with people, handover points, and changing transport tasks.
Optimization of reusable packaging cycles
MaaS-CU models reusable-packaging-as-a-service processes for the food industry so circulation counts, bottlenecks, and costs can be optimized with data.
We develop ML-based analyses for the product cycle, including digital-twin logic, bottleneck detection, outlier analysis, and decision-ready frontend views for operational users.
Personalized virtual try-on
mAImuse optimizes virtual clothing try-on through AI-supported image recognition and personalized product recommendations based on individual body shapes.
We define technical requirements, evaluate image, 3D, and virtual-try-on models, and integrate vision AI into a demonstrator for fit and appearance assessment.
Sensors are evaluated as a connected physical system so manipulated individual signals become visible in spatial context.
Intrusion detection from physical signal parameters
Physical Guards addresses IoT security in smart homes, production, and sensitive infrastructure using physical signal parameters instead of isolated sensor alerts.
We develop AI-based methods for multivariate time series, autoencoder-supported intrusion detection, and interpretable system fingerprints that reveal manipulation of individual sensors in the overall system context.
Transfer with public and industry partners
We support partners from feasibility studies through publicly funded collaborative projects to robust AI demonstrators.