Felix Friedrich
Hessian.AI and AI/ML Group, Computer Science Department, TU Darmstadt. Hochschulstrasse 1, Room S1|03 077, 64289 Darmstadt, Germany
friedrich (at) cs (dot) tu-darmstadt (dot) de


News. Please check out the latest news articles about our work at MIT technology review and TU Darmstadt website. Furthermore we are happy to showcase the newest demos on huggingface about our work: Fair Diffusion explorer and Sega demo.

Mission. My research interests are centered around artificial intelligence (AI) and how we can make such AI systems socially acceptable. For one, I focus on incorporating the human in the loop. Inspiration can come from cognitive science since the exploration and consideration of human behavior are inevitable and possibly the best motivation to improve current models. ChatGPT is just one example here where human interaction boosted performance and made people doubt that creativity is dead and AI is solved. While both are incorrect, it still demonstrates massive performance improvement through human interaction and the impact on current societies. On the other hand, we see that current models are heavily biased. This is not so surprising as they are trained weakly supervised on unfiltered data from the web and hence run the risk of reflecting or shifting these (web) biases. Therefore, I am also interested in better understanding the learned representations in these models and in how to debias them in arbitrary directions, paving the way for future AI models. To this end, I want to make use of human capabilities to incorporate knowledge outside of the rigid range of purely data-driven approaches.

Thesis/Praktikum (for CP) or Hiwi. We’re exploring several exciting projects aimed at improving the robustness of generative AI models, including large language models (LLMs), vision-language models (VLMs), and diffusion models (DMs). Our goal is to enhance their fairness, interpretability, and reliability. Most important is dedication and commitment; writing a thesis is a fulltime job. Many of my students end up being part of a published paper in great venues such as CVPR. On the other hand, if you look for a relaxed thesis, you might consider other supervisors. Furthermore, it is important that you have coding experience with PyTorch and Deep Learning in general and have experience with models from huggingface (like Stable Diffusion oder Llama3.X/GPT-4). If you have ideas yourself I'm also happy to discuss them. Please feel free to contact me in case of interest. Note: thesis proposal.

Timeline.
2021 - now: Member of 3AI at Hessian.AI (personal site).
2021 - now: Ph.D. student at the Machine Learning Lab, CS Department, TU Darmstadt, Germany.
2020: Erasmus+ at Chalmers University of Technology in Gothenburg, Sweden.
2019 - 2021: M.Sc. in computer science with minor in psychology from TU Darmstadt, Germany.
2018 - 2021: M.Sc. (with honors) in autonomous systems from TU Darmstadt, Germany.
2017: Research internship on intelligent autonomous driving systems at IAV, Germany.
2014 - 2017: B.Sc. in electrical engineering from TU Dortmund, Germany.

Supervised Courses.
WS 2024 Prof. Dr. Kristian Kersting, Probabilistic Graphical Models
WS 2023 Prof. Dr. Kristian Kersting, Probabilistic Graphical Models
SS 2022 Prof. Dr. Kristian Kersting, Data Mining and Machine Learning
WS 2021 Prof. Dr. Kristian Kersting, Introduction to AI
SS 2021 Prof. Dr. Kristian Kersting, Deep Learning: Architectures and Methods
SS 2020 Prof. Dr. Kristian Kersting, Statistical Machine Learning


Publications


Publications can be found at DBLP, GoogleScholar, SemanticScholar