Recommender Systems / ChessTEMPO
A recommendation model that fine-tunes itself to each player's memory, learning to act as a personal guide to their opening repertoire.
View projectAI Team Projects connect academic research with real-world AI and ML development. In this semester-long program, mixed international teams from Clausthal University of Technology and Babes-Bolyai University build practical intelligent systems together.
WORK HARD 💪 PLAY HARD 🏀
The absolute highlight of the ATP experience are the two onsite travel weeks, one hosted in Cluj and one in Goslar. These weeks are full immersion: students do not just hack together on AI models, they join all-day events where focused coding sessions are balanced with teambuilding, cultural exchange, and evening socials. The best technical ideas come from teams that know how to connect, adapt, and have fun together.
Pushing the boundaries of what's possible in cognitive systems and automated intelligence.
Recommender Systems / ChessA recommendation model that fine-tunes itself to each player's memory, learning to act as a personal guide to their opening repertoire.
View project
AI Team / Cognitive RoboticsA vision-based system combining person tracking, gaze, head and body orientation, motion, and gestures to estimate whether nearby people appear aware of a mobile robot.
View project
LLMs / Learning TechnologyA planned learning-platform demonstrator that keeps course material visible beside a context-aware LLM assistant and adapts support from a lightweight topic-level learning-state model.
View projectReserved for an upcoming AI Team Project. Details will follow when the scope and materials are ready.
Integrated web platform for IFC extraction, BIM question answering, XGBoost house-price prediction, room-image classification, and Stable Diffusion layout generation.
View ProjectCost-aware tender data extraction across changing procurement portals, combining reusable scrapers, browser automation, AI-generated extraction, self-healing retries, and document processing.
View Project
The Village at Scale trains LLM agents to reason and negotiate in Werewolf. It combines supervised fine-tuning from clean gameplay with reinforcement learning through self-play, while a live visualiser streams each six-agent match.
View ProjectNext.js dashboard that queries cluster nodes over SSH, parses nvidia-smi and Slurm data, and displays GPU health, running jobs, storage, training metrics, and benchmark history.
View ProjectPython benchmark that runs Ollama or Hugging Face models against each other in Stratego and logs prompts, moves, invalid actions, timing, outcomes, and player-position effects.
View ProjectA 1:10 vehicle testbed that combines SLAM, perception, path planning, feedback control, and autonomous parking in a compact end-to-end autonomy stack.
View ProjectUnity editor for assembling road layouts from reusable pieces, exporting OpenDRIVE 1.4 files, and loading the generated maps into CARLA with Python helper scripts.
View ProjectLooking for earlier iterations? These projects continue the tradition of our team projects from the University of Mannheim, where the same format ran under the European Master Team Project.
View previous projects at Uni Mannheim