Chair of Software Engineering for Business Information Systems (sebis)​
School of Computation, Information and Technology (CIT)​
​
Technische Universität München (TUM)​
Boltzmannstraße 3,​
85748 Garching, Germany

Phone +49 89 289 17114
Email tristan.till@tum.de
Room FMI 01.12.060
Office hours: by appointment

_{^{Curriculum Vitae}} 

Tristan Till has been a Research Associate at the Chair of Software Engineering for Business Information Systems at the Technical University of Munich since September 2025. He holds a Master’s degree in Artificial Intelligence from Nanyang Technological University, Singapore, where he wrote his thesis on enhancing the consistency of weakly-supervised medical image segmentation models. Beyond his studies, Tristan gained extensive research experience in Medical AI as a student research assistant at the Medical University of Graz. In early 2025, he also co-founded his own start-up, developing AI systems at the intersection of medicine and technology, including solutions for intraoperative guidance through surgical landmark detection, injury prevention in professional sports, and LLM-based child safety assistants.

_{^{Research Interests}}

• AI applications for medicine and healthcare
• Theoretical deep learning
• Real-time LLMs, Interaction Models and JEPA-style models
• Modern Information Retrieval

_{^{Open Thesis Topics}}

I offer thesis supervision for motivated master's students, often closely to my research interests but open for a wide range of topics in Deep Learning. If you would like to pursue a topic related to my research interests, please send me an email with your CV and transcript attached.

_{^{Extracurricular Research Experience}}

I am happy to collaborate with students looking for extracurricular research experience. Topics I am, have or will work on are:

• Limitations of spectral compressibility via transform-then-truncate schemes in deep learning
• Information content density shapes optimal decisions for agentic search
• A domain-specific language (DSL) for agentic retrieval pipelines
• Understanding ungrounded self-correction via tool-calls in long-horizon LLM agents via noise
• Measuring alignment via mental model completeness in interaction models
• Validating LLMs as simulators for interaction-model benchmarks
• Quantifying the real-time bottleneck of information flow in intelligent systems
• Bias in LLMs technically does not exist (but has very real implications)
• Measuring ESG disclosure from impact investment funds public data