AI Technologies projects from the Master’s program AI Engineering

Projects 2025

A small selection of the latest projects from the AI Engineering master’s program:

Cognitive Architecture for Industrial Knowledge Retention: An LLM-Based Agentic Memory Framework

Project: Sarah Eschenbacher
Category: AI
Description: This project is developing an AI-supported system for securing and passing on industrial expertise—especially in the context of a shortage of skilled workers. It combines information from written documentation with the experiential knowledge of employees. It uses an agentic memory model that employs four types of memory inspired by human memory. The aim is to make the transfer of knowledge between humans and AI assistants sustainable.

Audio-Based Classification and segment Detection of
RADIO broadcasts using machine learning tools

Project: Herbert Grünsteidl
Category: AI
Description: This project is developing an AI-supported system for securing and passing on industrial expertise—especially in the context of a shortage of skilled workers. It combines information from written documentation with the experiential knowledge of employees. It uses an agentic memory model that employs four types of memory inspired by human memory. The aim is to make the transfer of knowledge between humans and AI assistants sustainable.

Color Classification of Football Jerseys using Neural Networks

Project: Elias Lord
Category: AI
Description: Soccer is fast, dynamic, and colorful. In this project, artificial intelligence was trained to recognize jersey colors in real soccer videos, even in difficult lighting conditions or camera settings. The AI analyzes the video material in real time and correctly assigns the players. This opens up an innovative approach for coaches and analysts to evaluate game progress more efficiently and automatically in the future.

AI Chatbot

Project: Miloš Jovanović
Category: AI
Description: The project is developing an AI-powered search system that helps students quickly find reliable information in official university documents. The chatbot understands questions in German and English and provides direct excerpts from the original documents—including source links—without any hallucinations. The solution is open source, resource-efficient, and can even be run on standard laptops. This makes it an accessible entry point for using artificial intelligence in higher education.

Projects 2024

A small selection of the latest projects from the AI Engineering master’s program:

midas

Team: Alexander Pallisch, Momen Abdou, Adnan Vatric
Category: AI
Description: Stock market analysis: As part of the development project, an AI trading bot was programmed in Python that is connected to MetaTrader 5 and can trade completely independently on the forex and stock markets. This bot integrates comprehensive risk management and follows a moving average (MA) trend strategy. In addition, we have developed a neural network that analyzes financial values and can make predictions. This network is an important part of the bot and improves its trading decisions through professional forecasts.

OnHands

Team: Bushra Yasin, Alexander Kreissl, Philipp Ehm, Michael Schwingshackl, Marvin Elias
Category: AI
Description: More and more people want to travel and discover the world. Our luggage is our most faithful companion on all these journeys. Unfortunately, there are increasing numbers of lost suitcases and bags, which not only frustrates travelers but also puts a strain on Austrian Airlines staff. Manually entering this data can take over a minute per piece of luggage. In the OnHands project, we use computer vision and machine learning to identify the type and color of a piece of luggage scanned with a smartphone in a matter of seconds. This information is automatically entered into an international database for lost luggage.

Ultra Vision
(Sieger Showcase Evening: AI – Master 2024)

Team: Paul Hassanpour, Manuel Kranzl, Saifur Rahmani, Michael Zeiner
Category: AI
Description: The goal of this project was to develop an application for identifying faulty ultrasound transducers using test images. With a focus on implementing a binary image classification model, the tasks included data exploration, file standardization, and extraction of relevant image regions using deterministic and CNN-based methods. A model based on VGG16 was selected for classification. Preprocessing and data augmentation were explored but showed limited effectiveness. Hyperparameter tuning guided model optimization. The final application, implemented in Python and converted to an executable file, deals with the identification of defective ultrasound probes.

BelAIcik

Team: Gerald Kloimstein, Peter Szabo, Thomas Ville, Alexander Zotter
Category: AI
Description: BelAIcik is video analysis software designed to help coaches and scouts automatically analyze low-quality footage from a single angle. It includes several AI components for field recognition and player recognition and uses classic computer vision approaches for the rest. BelAIcik transforms side views into bird’s-eye views and calculates some statistics along the way. The prototype has shown that some information can be extracted even from a video view.

Projects 2023

A small selection of the latest projects from the AI Engineering master’s program:

Content CreAItors (Sieger Best of Class Award 2023)

Developers:

The aim of the Content CreAItors project is to train a natural language model so that it can generate a suitable YouTube title from a summary of a text. The focus was limited to the topic of edutainment. The summary is generated from the video’s subtitles, which allows the entire process to be automated.

League of Bots

Developers:

The League of Bots project focused on the development of a 2D platform game in which four artificial intelligences compete against each other. For this purpose, a dynamic level was created, consisting of various handcrafted level segments. Parameters were added to train the AIs to randomly generate levels for training. Challenges during the semester were used to determine the best AI. Human players can now compete against the AI in predefined levels.

Der Blickrechner /  Sum at first sight

Developers:

With the help of the software Der Blickrechner, a glance is all it takes to make PCs accessible. For many people with disabilities, things that we take for granted are often impossible, such as controlling a computer with a mouse. In this project, the mouse does not have to be controlled by hand. An example is a calculator app that allows you to add, subtract, multiply, and divide—using only your eyes.

Language Sloth

Developers:

The Language Sloth project is a native application that deals with speech synthesis and translation of texts and voice recordings. Speech synthesis is the artificial generation of the human voice. The application has two functions. The first function is speech-to-speech, which allows users to record speech using the application and select the language into which the recorded text should be translated. The result can then be played back directly in the application. The second function is a text-to-text translator. As the name suggests, users can enter text into the application, which is then translated into the desired language.

Pick and Place with AI (PnP w/ AI)

Developers:

Pick and Place with AI’s AI recognizes in real time which components are currently being captured by the camera. The accuracy of the component recognition is displayed on the screen for each component as a percentage. Current automatic placement machines perform their task very well, but their price is a major disadvantage. Instead of purchasing a machine for thousands of dollars, this project focuses on training an AI model that predicts the position and rotation of electrical components, such as computer chips, in an image. The model is trained exclusively with synthetic data. The AI of Pick and Place with AI recognizes in real time which components are currently being captured by the camera. The accuracy of the component recognition is displayed on the screen for each component as a percentage.