LUCA KOHLHEPP: ML driven RFI filtering Context: Radio interferometry is an important method in astronomy to archive high resolution images, using multiple telescopes. This method is susceptible to disruptions, which degrade image quality dramatically so-called radio frequency interference (rfi). For proper imaging this rfi needs to be flagged to be excluded for the imaging process. Aims: Development of a machine learning algorithm that can successfully flag rfi of GMRT observations, to a degree, where imaging with it is possible. Methods: For this an CNN is used. Which uses 3D-convolution layers. Therefore, observation time is used as one of the dimensions for convolution. The ground truth for supervised learning is calculated by the established tfcrop algorithm. Results: The CNN trains successfully on the data. The best results are archived only a single convolution in the time-axis.