Our brain needs to constantly fuse sensory information detected by our multiple senses to produce a seamless coherent representation of the world. Rather than being the exception, this binding process is ubiquitous to sensory-motor integration and is implicated in most cognitive functions. Its impairment is a cause of various pathologies, such as schizophrenia or autism. Multisensory processing operates on all brain levels from primary cortices over subcortical structures up to higher associative centers, while the smallest operational units are single multisensory neurons. In an interdisciplinary effort, we combine optical developments, genetics, and neuro-computation to obtain new insights into the activity of brain-wide neural circuits that process multisensory information. To reduce the complexity, we study the small transparent brain of zebrafish larvae as a model system. Our research particularly examines gaze stabilization, a multisensory task inherent to all vertebrates. This reflex integrates vestibular and visual inputs to control eye movements, compensating for self-motion and ensuring visual clarity. In this project, we have developed a series of innovative experimental platforms that allow for the integration of vestibular stimulation with high-resolution, brain-wide neural activity recordings. Our pioneering efforts have mapped the vertebrate brain's global response to vestibular stimuli related to roll and pitch movements. By combining vestibular and visual stimuli in various multisensory scenarios—coherent, asynchronous, or conflicting—we have identified multisensory neurons and assessed their functional significance across the brain. Through in silico studies employing minimal spiking neural network models, we have examined the roles of multisensory neurons in a variety of tasks. Our findings suggest that while multimodal units may not be necessary for achieving accuracy or managing the speed-accuracy trade-off in traditional multisensory tasks, they are indispensable for tasks requiring the detection of comodulation across sensory channels. Additionally, our work has led to the deduction of brain-wide functional connectivity, utilizing generative network modeling based on recorded brain activity. With the introduction of an innovative all-optical platform for optogenetic circuit interrogation, our ongoing work is aimed at further delineating these multisensory networks and confirming computational predictions emanating from our models.