The input-output transformation of individual neurons is a key building block of neural circuit dynamics. While previous models of this transformation vary widely in their complexity, they all describe the under-lying functional architecture as unitary, performing a single type of elementary computation, either once, or cascaded multiple times. Here, we show that the input-output transformation of CA1 pyramidal cells during phase precession in the ‘theta-state’ is instead best captured by two distinct functional architectures operating in parallel. We use statistically principled methods to fit flexible, yet interpretable cascade mod-els of the transformation of the spatiotemporal patterns of input spikes into the somatic ‘output’ voltage, and to automatically select among alternative functional architectures. For this, we first extend previous methods that only included subunits with static nonlinearities by incorporating subunits expressing arbitrary nonlinear dynamics. Second, we develop methods to fit all the parameters of the model, including those parameterizing the dynamical nonlinearities, as well as the architecture of the model in a data-driven way, with minimal prior assumptions. We find that predicting the contribution of dendritic Na+ spikes (vNa) and all other dendritic signals (v_other) to the output of the cell requires two, fundamentally distinct functional architectures. Specifically, while v_other can be accurately predicted using a single subunit with a static nonlinearity, precisely capturing v_Na requires several dendritic subunits equipped with nonlinear dynamics and connected into an architecture that appropriately reflects the clustering of synaptic inputs onto the dendritic tree. Moreover, automatic architecture-discovery reveals that the v_other-architecture reflects the somatic distance of synapses, while the v_Na-architecture reflects the clustering of synapses on the dendritic tree. The presence of two distinct, parallelized functional architectures within each individual neuron suggests potentially far-reaching consequences for our understanding of the dynamics of cortical circuits.