ICE Industrial-AI 🧠🤖
Applied Artificial Intelligence
ICE Industrial-AI
Research & Developement Project Partner
Welcome to ICE Industrial-AI — as part of the OST Institute of Computation Engineering ICE the Industrial-AI team focuses on applied industrial artificial intelligence. We bridge cutting-edge AI technologies with real-world industrial challenges to drive innovation, efficiency, and performance. From concept to deployment, we collaborate with industry to create intelligent, future-ready solutions.
Latest from the Blog
ML Techniques in Defensive and Offensive Cybersecurity

ML Techniques in Defensive and Offensive Cybersecurity
Latest

The integration of Artificial Intelligence into cybersecurity has created a dual-use dynamic. It is a tool of immense power for defenders (Blue Teams) seeking to automate threat detection, yet it simultaneously empowers attackers (Red Teams) to generate sophisticated, polymorphic threats. This article explores the spectrum of machine learning techniques—from standard supervised learning to cutting-edge Generative AI—and how they are applied on both sides of the digital battlefield.
Vulnerabilities of Deployed Machine Learning Models

Vulnerabilities of Deployed Machine Learning Models
Latest

As Artificial Intelligence systems move from research labs to critical deployment in sectors like automotive, healthcare, and finance, their security profile changes drastically. This article explores the specific vulnerabilities of deployed machine learning models, ranging from mathematical adversarial examples to semantic attacks on Large Language Models (LLMs). We analyze the attack surface across the ML lifecycle and discuss strategic defenses.
From Vectors to Graphs: The Evolution of Retrieval-Augmented Generation

From Vectors to Graphs: The Evolution of Retrieval-Augmented Generation
Latest

Standard Retrieval-Augmented Generation (RAG) systems rely heavily on vector embeddings to retrieve relevant context. While effective for local semantic queries, these systems often fail to capture global structural information and multi-hop relationships across large corpora. This article introduces Graph RAG, a paradigm shift that integrates Knowledge Graphs (KGs) into the retrieval pipeline. We explore the architectural differences, the mathematical formulation of graph-based retrieval, and the Community Summary approach for global reasoning.
Latest Events
AI-Stammtisch 6.25

AI-Stammtisch 6.25
Latest

AI-basierte Optimierung von Metall-Membranbälgen
AI-Stammtisch 5.25

AI-Stammtisch 5.25

Ergotherapie & AI: Mit Hilfe von AI zurĂĽck in den Alltag
AI-Stammtisch 3.25

AI-Stammtisch 3.25

Large Language Models for Software Engineering
Latest Projects
Know the flow! Unique Footfall Information as USP in Payment and PropTech

Know the flow! Unique Footfall Information as USP in Payment and PropTech
Finished

Development of a footfall model fed by a unique set of real data in order to generate footfall information in an unknown granularity (time and location), on-demand and as a basis for an ecosystem of location-intelligent products, and services for both implementation partners.

AI
Autonomous Drone System for Precise Localization of Avalanche Victim

Autonomous Drone System for Precise Localization of Avalanche Victim
Finished

Every year, around 100 people die in avalanche accidents in Europe alone. Together with industry partner ANAVIA from Näfels GL, a system is being developed to localise avalanche victims as quickly as possible. The project idea was selected as the winner of the 5th Innovation Boosters Robotics call alongside five other projects.

AI Signal Processing
Latest Publications
Key Vital Signs Monitor Based on MIMO Radar

Key Vital Signs Monitor Based on MIMO Radar
Accepted

Research on advanced optimization techniques for improving efficiency in smart manufacturing environments.

Computer-Aided diagnosis Convolutional Neural Networks Displacement Measurements Machine Learning MIMO Radar Sleep Apnea
Parameter-free Neural Field-based Optimal Design of Nonuniform Transmission Lines

Parameter-free Neural Field-based Optimal Design of Nonuniform Transmission Lines
Conference

This paper presents a novel method for resolution free, free-form shape optimization for nonuniform transmission lines using artificial neural networks. The research introduces an innovative approach that leverages machine learning techniques to optimize transmission line designs without being constrained by traditional parametric limitations..

AI Electronic Design Automation Deep Learning Geometric Representation Learning