Christian Limberg

WORK HISTORY

EDUCATION

ABOUT CHRISTIAN LIMBERG

I am an enthusiast in applied and explainable AI, driven by a vision to find intelligent solutions for tackling today\'s major problems and improving company processes with the most recent technologies. During my research, I applied Machine Learning to solve complex Computer Vision and Robotics challenges. Since my master\'s studies in 2013, I focused extensively on these areas, mastering numerous skills, tools, and APIs. I e.g. participated in the RoboCup@Home competition with my university team, which taught me early on that seamlessly integrating a Machine Learning model into human environments involves much more than merely deploying it on a robot. This realization guided my PhD research in collaboration with the Honda Research Institute EU, where I developed meta Machine Learning approaches for robot self-assessment in human-robot teaching scenarios to ensure reliability and effectiveness. In my postdoctoral research at the National Institute of Informatics (NII) in Tokyo, I extended my expertise in Generative AI, specifically Transformer-based models, which are the foundation of approaches like GPT-4. I explored Generative AI for audio by training a latent 2D audio map with a Variational Autoencoder to condition a GPT-2 model for more natural audio generation. Additionally, for Vision-Language Models, I combined YOLO World for detecting people and GPT-4V for classifying their actions to create a flexible perception framework for autonomous drones. My current work on this topic involves investigating the trustworthiness of GPT-4 explanations and their impact on classification accuracy.