Studying various properties of stellar populations within a galaxy provides rich insight into its formation, and current surveys enable us to do so with the goal of reconstructing their formation histories. However, it is ambiguous as to when the main progenitor of a galaxy like the Milky Way (MW) formed, and what its constituent mergers were like. In the first half of this work, I use the FIRE-2 cosmological simulations to study these low-mass galaxies at early times, and their later mergers. I quantify the formation times of MW/M31-mass galaxies based on different definitions of formation, as well as the mass function of its constituent building blocks. I find that the MW-mass galaxies in Local Group (LG)-like pairs form earlier, highlighting the role that environment may play in galaxy formation. I also find a similar ubiquitous feature of metal-poor stars in the disks of the MW-mass hosts on preferentially prograde orbits, which are also found in the MW. These stars largely originate ex-situ, from a gas-rich merger of mass similar to the Magellanic Clouds, that sometimes shapes the orientation of the stellar disk.

We call the gravitationally bound low-mass galaxies that survive the hierarchical formation process satellite galaxies. The satellite galaxies in the LG are the only low-mass galaxies that we can extensively study through fully resolved stellar populations and their full phase-space coordinates. In the second half of this work, I study the infall and orbit histories of satellite galaxies around the MW-mass hosts in the simulations. I examine trends in their present-day orbital dynamics, including total velocity, specific angular momentum, and total energy, as well as their orbital histories versus present-day distance, stellar mass, and the lookback time of infall. Surprisingly, the majority of satellites that experience multiple orbits have larger recent pericenters than their minimum, contradictory to the assumption that satellite orbits always shrink. Finally, I compare the satellite galaxy orbits in a static, axisymmetric MW-mass potential to the simulations to quantify the extent to which this common orbit modeling technique works. I find that orbital energy and specific angular momentum are not always conserved, and orbit properties that occurred recently, such as pericenters, are better recovered than events further in the past, such as infall time.