I developed an in silico model of mammalian hematopoiesis using mathematical modeling, machine learning, and high performance computing to better understand clinically relevant perturbations. Strict regulation by feedback cytokines allows the system to both maintain homeostasis and respond rapidly to perturbations. Because feedbacks introduce nonlinearities, training cannot be accomplished by calculating exact gradients and thus must be approximated by other means. In line with OpenAI’s findings, I found evolutionary strategies to be a successful and scalable method for approximating gradients in the large, complex biological systems our lab typically deals with. So I wrote a generalized evolutionary strategy framework to optimize parameters for this and other systems. Using my framework I am able to effectively search the large parameter space and arrive at a parameterized model of mammalian hematopoiesis that fits experimental data, makes insight into progenitor activity, and predicts clinically relevant pertubations such as rituximab induced late onset neutropenia.