Advanced technology development of industrial applications in safety critical systems using Machine-Learning.
-
Pattern Recognition in Time-Series using state-of-the-art Deep-Learning Methods
-
Data Preparation, Database Design and Creation of UI-Tools for the labeling of data (approx. 100k datasets.)
-
Creation of feature-engineered algorithms for the pre-labeling of the data
-
Research on current state-of-the-art methods (Scientific Papers)
-
Development of the Model-Architecture, proper modeling of Cost-Functions. Evaluation and Application of state-of-the-art Algorithms (such as U-Net, Inception, ResNet and YOLO) to other domains (i.e., various sensors in mechatronic applications)
-
Implementation of suitable Performance-Measures and Error Visualization
-
Planning and execution of Experiments, creation of an End-to-End Loop for Model-Verification by use of a slim CI/CD Pipeline. Creation of Docker-Images for use in AI-Infrastructure (cloud)
-
Filing of Invention-Disclosures, stakeholder communication, documentation
-
Optimization of the Model to make it fit on embedded Hardware
Tools used: Python (incl. Numpy, Pandas, Scikit-Learn, ..), TensorFlow, Keras, PyQt, Docker