As Austria's largest research and technology organisation for applied research, we are dedicated to make substantial contributions to solving the major challenges of our time, climate change and digitalisation. To achieve our goals, we rely on our specific research, development and technology competencies, which are the basis of our commitment to excellence in all areas. With our open culture of innovation and our motivated, international teams, we are working to position AIT as Austria's leading research institution at the highest international level and to make a positive contribution to the economy and society.

Our Center for Vision, Automation & Control (VAC) is a team of approximately 140 experts carrying out applied research in image processing, sensor-data fusion, machine learning, data analytics, mathematical modeling for real-time applications, process automation, and control of complex dynamic systems for industrial use. Within VAC, the Competence Unit Assistive & Autonomous Systems (AAS), located in Vienna, invites applications for a Master Thesis project. Our competence unit researches and develops technology components for assistance systems and the automation of commercial vehicles and unmanned aerial platforms, focusing on environment-perception techniques (e.g., multi-spectral cameras, laser- and radar-based sensors, odometry, and satellite navigation) to support efficient, flexible, and safe AI-driven assistance. VAC works closely with national and international scientific and industry partners to advance prototypes from laboratory demonstrations through to industrial deployment.

Throughout your master’s thesis you will contribute to the TeleAssist-FFG-Project, an initiative to transform industrial teleoperation into a truly collaborative experience. By combining live sensor and machine‐state feeds with embedded technical documentation in an LLM‐driven backend, TeleAssist doesn’t just let operators pilot machines remotely - it monitors for anomalies, interprets error codes, and proactively guides users through troubleshooting steps as they work. The goal of your thesis will be to develop and evaluate event-driven data abstractions that distill raw time-series into concise, meaningful tokens, enabling the LLM assistant to detect and explain faults in real time and empower even non-specialists to resolve errors safely and efficiently.

• Dive into the Literature: Conduct a comprehensive literature review on time-series abstraction, tokenization strategies, and LLM-mediated diagnostic workflows, and refine the thesis scope and crystallize clear research questions.
• Design and implement an Abstraction Library: Create a modular Python package implementing event-driven techniques (threshold alarms, change-point detection, windowed summaries, symbolic encoding, event clustering) that can be easily extended and integrated.
• Work with Diverse Data Sources: Process and integrate public benchmark datasets, ROS 2/Gazebo-simulated streams, and real industrial time-series to build realistic evaluation scenarios.
• Design & Execute an Evaluation Framework: Define and implement metrics (compression ratio, reconstruction fidelity, LLM diagnostic accuracy), craft effective prompt templates, integrate an open-source LLM, and perform systematic comparative experiments.

• Learn Scientific & Engineering Practices: Gain practical experience in scientific methodology, experimental design, data analysis workflows, version control, and code documentation. Participate in regular progress reviews and present interim findings to a panel of interdisciplinary experts.
• Prototype & Document Results: Collaborate with our researchers to refine your prototypes, interpret results, and generate actionable recommendations for best-practice abstraction strategies. Deliver production-quality code, reproducible experiments, and a polished thesis report.