Machine learning based long-lived particle reconstruction algorithm for Run 2 and upgrade LHCb trigger and a flexible software platform for the UT detector read-out chip emulation

This thesis discusses two projects, which were performed during the author's work for the LHCb Collaboration. The first one was related to the design and improvement of the algorithm dedicated to reconstructing long-lived particles, such as $K_{s}$ or $\Lambda$ hadrons. This algorithm is used to reconstruct the particles that decay outside of the Velo detector and depends on the input from the tracking station situated downstream from the Velo. Thus, it is called Downstream tracking. This reconstruction algorithm's time budget is minimal because it is executed as a part of a real-time LHCb trigger system. To significantly reduce a ghost stream, two machine learning classifiers were trained and deployed. The second part of this thesis was related to the design and implementation of the Upstream Tracker raw data emulation and performance monitoring software platform. This software was essential for processing the data collected during the testbeam experiments and obtained results were essential for a number of published papers by the UT grup. In the future, this software will be used to perform calibration of the UT processing algorithms as well as for performance diagnostics of the UT detector.