Gravity currents are buoyancy driven flows that appear in a variety of situations in nature as well as industrial applications. Typical examples include avalanches, oil spills, and turbidity currents. Most naturally occurring gravity currents are catastrophic in nature, and therefore there is a need to understand how these currents advance, the speeds they can attain, and the range they might cover. This dissertation will focus on the short and long term evolution of gravity currents initiated from a finite release. In particular, we will focus attention to hitherto unaddressed effect of the initial shape on the dynamics of gravity currents. A range of parameters is considered, which include the density ratio between the current and the ambient (heavy, light, and Boussinesq currents), the initial height aspect ratio (height/radius), different initial cross-sectional geometries (circular, rectangular, plus-shaped), a wide range of Reynolds numbers covering 4 orders of magnitude, as well as conservative scalar and non-conservative (particle-driven) currents. A large number of experiments have been conducted with the abovementioned parameters, some of these experiments were complemented with highly-resolved direct numerical simulations. The major outcome is that the shape of the spreading current, the speed of propagation, and the final deposition profile (for particle-driven currents) are significantly influenced by the initial geometry, displaying substantial azimuthal variation. Especially for the rectangular cases, the current propagates farther and deposits more particles along the initial minor axis of the rectangular cross section. This behavior pertaining to non-axisymmetric release is robust, in the sense that it is observed for the aforementioned range of parameters, but nonetheless cannot be predicted by current theoretical models such as the box model, which has been proven to work in the context of planar and axisymmetric releases. To that end, we put forth a simple analytical model (an extension to the classical box model), well suited for accurately capturing the evolution of finite volume gravity current releases with arbitrary initial shapes. We further investigate the dynamics of a gravity current resulting from a finite volume release on a sloping boundary where we observe some surprising features.