Master-Thesis “World Models for Physical AI”

AIT Austrian Institute of Technology; Wien; Hybrid; EUR 10.000 - 12.000; Vollzeit

Benefits dieser Stelle

Hybrid work

Mentorship

Publication potential

Flexible start

Zusammenfassung

The AIT Austrian Institute of Technology invites applications for an exciting master thesis in the Center for Vision, Automation & Control, Vienna. You will investigate how world models can be applied to field robotics using real-world LiDAR, image, and sensor data to enable efficient, AI-driven perception and control for large-scale machinery.

You will develop and compare encoder strategies for LiDAR and camera data and assess learning objectives; the project emphasizes on-device feasibility

Qualifikationen

• Ongoing master’s studies in AI, computer science, robotics, or a related field.
• Strong programming skills in Python and PyTorch.
• Experience with 3D data or point cloud processing is a plus.

Aufgaben

• Explore world models for physical AI with real-world LiDAR, image, and sensor data.
• Design efficient world model architectures for on-device AI in industrial settings.
• Implement encoders for LiDAR and image data (voxel, point-based, BEV, or tokenized).
• Evaluate predictive quality and robustness under challenging conditions.

Kenntnisse

Python

PyTorch

3D data

Point cloud

Computer vision

Deep learning

English

Ausbildung

Master’s studies in AI/CS/Robotics

Tools

Git

Jobbeschreibung

The AIT Austrian Institute of Technology is Austria’s largest Research and Technology Organisation (RTO). We aim to transform scientific excellence into innovation with impact. Together with our partners, we develop technologies and solutions for a sustainable, resilient, and digitalized future, strengthening Europe’s competitiveness and technological sovereignty. Our international teams combine cutting‑edge research with strong implementation expertise to create measurable value for industry and society.

The Center for Vision, Automation & Control located in Vienna (Giefinggasse, 1210) invites applications for a master thesis. Within the Center for Vision, Automation & Control, the Competence Unit Assistive & Autonomous Systems (AAS) 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.

As part of this master's thesis, you will investigate how world models can be brought to field robotics using real‑world‑scale LiDAR, image, and sensor data. The goal is to enable physical AI for large‑scale machinery, such as cranes, by designing efficient world model architecture that run on on‑device hardware. Your work will be embedded in the LOK (Logistik‑Onboard‑Kompetenz) project, which aims to modernize intermodal transport logistics through forward‑looking, on‑device AI solutions for the efficient handling of transport containers. LOK's central challenge is resolving the trade‑off between accuracy and speed, so that novel AI architectures and multi‑sensor setups remain reliable even in complex scenarios and adverse environmental conditions.

This is your opportunity to tackle one of today’s hottest AI topics, deliver immediate real‑world impact in industrial maintenance, and supercharge your career in AI and data science.

CENTER FOR VISION, AUTOMATION & CONTROL

• Under the guidance of our expert team, you will immerse yourself in both the theoretical and practical aspects of world models for physical AI. Throughout the thesis, you’ll receive hands‑on mentorship, structured learning sessions, and regular interdisciplinary feedback to ensure your professional growth and project success.
• Dive into the Literature: Review world model architectures for physical AI, with particular attention to the contrast between reconstruction‑based approaches and JEPA‑style latent prediction, as well as self‑supervised representation learning on large‑scale LiDAR and multi‑sensor data.
• Encoder Strategies: Implement and compare encoder designs for LiDAR and image data, e.g., voxel‑based, point‑based, bird's‑eye‑view, or tokenized representations, and assess how well they suit world model learning.
• Learning Scene Dynamics: Design and train world models that capture how the scene evolves over time and systematically compare learning objectives and architectural choices.
• Evaluation: Define suitable metrics and evaluate predictive quality and robustness under challenging conditions, including the transfer gap between simulated and real sensor data.
• On‑Device Feasibility: Explore the compatibility of the developed approaches with edge and on‑device hardware.
• Results & Publication: Analyze and present your findings, derive recommendations for future research, and, depending on your results, contribute to a publication at a leading robotics or machine learning venue.

Your Qualifications As An Ingenious Partner

• Ongoing master’s studies in Artificial Intelligence, Computer Science, Robotics, or a comparable technical field.
• Strong programming skills, especially in Python and PyTorch, along with a solid understanding of deep learning; experience with 3D data or point cloud processing, computer vision, or generative models is a plus.
• Enjoyment of application‑oriented challenges in industrial contexts, translating theory into real‑world solutions.
• High level of commitment and team spirit, with the ability to collaborate effectively in multidisciplinary settings.
• Excellent English skills, both written and spoken, for scientific communication and documentation. German skills are a plus.
• Academic supervision support: If you already have a fitting supervisor at your university, that’s ideal; otherwise, we’ll help you connect with an appropriate advisor.

What To Expect

• Duration: 6 months (flexible start, ideally as soon as possible)
• Location: Vienna (hybrid)
• Intensive mentorship by experienced AIT scientists and experts
• Opportunity to publish results in peer‑reviewed conferences and collaborate with leading industry partners
• EUR 1027,40 gross per month for 20 hours/week based on the collective agreement. There will be additional company benefits. As a research institution, we are familiar with the supervision and execution of master theses, and we are looking forward to supporting you accordingly!

At AIT, we create an inclusive and family‑friendly working environment that promotes equal opportunities and actively strengthens diversity across our workforce and in leadership positions. Among other initiatives, we are committed to increasing the proportion of women in our company and therefore particularly welcome applications from female applicants.