This project requests computational resources to support the scalable analysis of large-scale in vivo neuroscience datasets generated across multiple experimental paradigms. Our overarching goal is to understand how distributed brain circuits give rise to behavior by integrating high-density neural recordings with detailed, time-resolved measurements of animal behavior. The datasets primarily consist of Neuropixels electrophysiology recordings synchronized with high-speed video (200–300 Hz) and other behavioral measurements. Neuropixels probes simultaneously record electrical activity from hundreds to thousands of neurons across multiple brain regions using thousands of densely packed recording sites, producing large raw data streams that require computationally intensive preprocessing steps such as spike sorting, quality control, and alignment across modalities. Experimental paradigms include decision-making tasks with simultaneous recordings from multiple brain regions, large-cohort studies across diverse transgenic mouse models relevant to neurodevelopmental disorders, and virtual-reality experiments in which neural activity is recorded during continuous, self-generated movement. These experiments are conducted by multiple experimentalists across several independent setups and involve recordings from many animals, resulting in datasets that are large in volume, multimodal in nature, and continuously expanding. Our laboratory has extensive experience in acquiring, preprocessing, and analyzing these data using established computational pipelines. However, the increasing scale and complexity of the datasets—driven by high-channel-count electrophysiology, high-speed video, and large experimental cohorts—now exceed the capacity of local computing resources. The planned analyses rely heavily on GPU-accelerated workflows, including AI-based tools for behavioral quantification and neural data analysis that run in CUDA-enabled environments. Access to NAISS computational infrastructure is therefore essential to enable efficient parallel processing, GPU-based model training and inference, and large-scale cross-dataset analyses. These resources will allow us to scale existing, validated analyses to larger datasets and fully exploit the richness of the data to address our scientific questions.