Many of the natural and mathematical sciences rely on differential equations to model dynamical systems--for example, atmospheric dynamics, population dynamics, or, as will be discussed in this talk, the dynamics of cellular networks. Moreover, we increasingly have access to large amounts of data describing the dynamics of these systems. Yet, the problem of learning differential equation models from such data is greatly understudied in the machine learning and statistics communities. This talk will describe a particular application of differential equation learning targeted at the segmentation gene network of fruit flies, a canonical example of a developmental gene network. Along the way, I will make general comments about the computational difficulty of the learning problem in its usual formulation. I will also present a trick for reformulating the problem that both makes it much easier to solve and allows us to utilize many existing machine learning techniques in its solution.