Data-centric Artificial Intelligence

  • Background

    Machine learning is the driving force behind most progress in the field of Artificial Intelligence. In the last decades research about machine learning focused on increasing the accuracy of models by adapting learning algorithms, neural network architectures etc. To this end, datasets are held constant – both on competition platforms like Kaggle as well as the benchmark datasets used for scientific publications. As a result of this approach machine learning models reached a high level of maturity and accuracy.

    In real-world problems datasets (e.g. images or sensor time series from production lines) are typically less curated than established benchmark datasets. The old saying “garbage in, garbage out” is especially true for machine learning. In more academic terms the data quality significantly determines the results of machine learning models. Based on this understanding a new approach to machine learning emerges: Data-Centric Artificial Intelligence. Several publications have already shown that this approach can yield significant accuracy gains and thus is an important alternative and complementary approach to tweaking model parameters & architectures.


    Research Goal

    We want to structure the field of Data-Centric AI and implement and evaluate existing approaches on data-centric AI in general and instance importance in specific.

    Working on this thesis you will:

     - Dive into an exciting and promising new field of Artificial Intelligence

     - Make your great theoretical skills even more practically relevant 

     - Collaborate with motivated and experienced researchers and practitioners

    We look forward to receiving your application because you:

     - have solid programming skills in Python (e.g., pytorch, tensorflow, keras)

     - have experience with state-of-the-art computer vision and deep learning frameworks (e.g., R-CNN, ResNet, AlexNet)—or are willing to learn quickly

     - want to gain first-hand experience solving real-world problems


     - Start: immediately 

     - Duration: 6 months

     - Location: up to you

    We offer you a challenging research topic, close supervision, and the opportunity to develop practical and theoretical skills. If you are interested, send your CV, transcript of records, and a brief letter of motivation to