In this seminar talk, I will present the current status of in-ice radio detection of ultra-high-energy (UHE, E > 10^16 eV) neutrinos and how we use deep learning to improve the trigger system and event reconstruction. UHE neutrinos are connected to the most energetic phenomena in our universe and neutrino astronomy is a powerful tool to study the high-energy universe. Their small flux and cross-section require the instrumentation of huge volumes which can be achieved with a sparse array of radio detector stations installed in polar ice sheets. I will report on the results from the pilot arrays ARA and ARIANNA, the status of the Radio Neutrino Observatory in Greenland (RNO-G) that is currently being deployed and the plans for IceCube-Gen2.
In the second part, I will talk about the usage of deep learning techniques. I will present a pilot study that uses a small convolutional neural network for a real-time rejection of thermal noise which allows to lower the trigger thresholds and leads to an increase in neutrino sensitivity by up to a factor of two. Furthermore, I will present reconstruction algorithms for the neutrino direction, energy and flavor based on deep neural networks, that – for the first time – allowed to quantify the sensitivity to the neutrino flavor.