The Standard Model (SM) of particle physics is the established theory of fundamental particles and their interactions. However, there are open questions that the SM does not account for, thus the search for New Physics (NP) by observing signals deviating from the predictions of the SM is an important area of experimental physics. The particle detectors of Large Hadron Collider provide a window to search for particles produced in proton-proton collisions, where a promising proxy for NP signals is the jet, the collimated energy shower deposited by hadronized partons in the calorimeters of the detector. Jet identification can be formed as a supervised machine learning classification task but constructing dedicated taggers is a laborious task. This problem motivates the creation of unsupervised anomaly detection models, trained with the known SM background, to provide an estimation for unknown signals to originate from NP particles, which is the main subject of this master thesis.