Gaussian Process (GP) models have gained popularity for its flexibility to handle correlation among data sampled from distributions in the exponential family. The correlation frequently characterizes time-dependent data, such as step count data across different time horizons. In this project, we plan to analyze the effectiveness of Gaussian Process on zero-inflated Poisson (ZIP) step count data. To do so, we evaluate the performance of Gaussian Process in fitting a series of generative models approaching the ZIP distribution. Our approach is to create a switched likelihood function in our Gaussian Process with the switched parameter a proxy of our $\pi$ variable denoting the probability of zero-inflation. A flexible GP should be able to successfully model any mixture of generative distributions from the exponential family and zero-inflated data. By being able to model the underlying generative distribution, we can better identify the time-varying treatment effect of mobile health interventions for step count data.