As of 2026, the Square Kilometer Array (SKA) will start to generate astronomical data. In order for the astronomical community to get used to and familiar with the type of data that can be expected from the SKA, the SKA-organisation generates so-called SKA Science Data Challenges (SDCs). In 2021, the Onsala Space Observatory (OSO) entered the second incarnation of these competitions, SDC2. In conjunction with a team from Fraunhofer-Chalmers, OSO developed a machine-learning based algorithm to find and characterise emission from neutral Hydrogen (HI) in a simulated data cube roughly 1TB in size. The algorithm was developed and run on computing resources at the Swiss National Supercomputing Centre (CSCS) in Switzerland. The team became second in a competition involving 6 participants. In the predecessor of this proposal, we successfully built and deployed the same package on the Alvis cluster. In this round, we would like to 1) test its performance on both the data from SDC2 and real data from the Low Frequency Array (LOFAR) and APERTIF (APERture Tile In Focus on the Westerbork radio telescope), this might involve retraining the model; and 2) adapt the code to also work on magnetism data for the upcoming SDC4.