Molecular Dynamics (MD) is a powerful computing tool that can give us insight into how bio-molecules interact without doing hands-on experiments. However, MD is limited by the fact that current computing resources can only simulate a handful of molecules on a microsecond timescale. Hence, a coarse-grained model with reduced degrees of freedom can be useful, since it can give us the same useful information with orders of magnitude reduced computational time. We present a coarse-grained model based on the Mori-Zwanzig formalism. Our approach leads to the computation of a rigorously correct generalized Langevin equation (GLE), which reproduces the correct kinetics. As an example, we show that the model reproduces the distribution of first passage times and velocity correlation functions for alanine dipeptide. In addition, we show that the memory part of the GLE is essential to reproduce the correct kinetics. In other words, in our example, Markovian models fail to reproduce brute force MD results, whereas the GLE succeeds. Finally, we apply our model to bigger bio-systems to measure rates of rare events. Specifically, we will present results for the rate of flip-flop (80 degree orientation change) of protonated oleic acid molecules in a phospholipid bilayer.