LISA is a space-based gravitational wave (GW) observatory that is planned for launch in 2034. It consists of three satellites in the free fall in the heliocentric orbit forming an equilateral triangle. Satellites exchange the laser light forming transponding interferometry allowing to detect GWs in the mHz band. A plethora of gravitational-wave signals from different astrophysical sources is expected to be observed by LISA. These sources include galactic white dwarf binaries, extreme mass-ratio inspirals, massive black hole binaries, etc. These signals could be very loud and remain in the LISA data for a long time (from days to years, depending on the source). Disentangling and characterizing these signals is crucial for further inference of astrophysical models for the formation and evolution of the population of GW sources. This project consists of the development of data analysis techniques able to disentangle GW signals in LISA data. The catalogues of the inferred source from the simulated data will be used in reconstructing the astrophysical model(s) used in the simulation (inverse problem).