How Kodiak trains the brain behind 28 driverless trucks
Twenty-eight trucks, and no humans in the cab. As of March 31, 2026, Kodiak's autonomous driving system, the Kodiak Driver, runs commercial freight on public roads across long-haul trucking, and industrial applications. This is the forefront of ground autonomy. At every mile, Kodiak’s core value...
Twenty-eight trucks, and no humans in the cab. As of March 31, 2026, Kodiak's autonomous driving system, the Kodiak Driver, runs commercial freight on public roads across long-haul trucking, and industrial applications. This is the forefront of ground autonomy. At every mile, Kodiak’s core value proposition rests in setting new standards for safe and reliable freight hauling that reshape the road ahead. The system behind it is GigaFusionNet, powering their autonomous driving system. Autonomous driving at the level above human competency and safety demands a paradigm shift in how we build AI. GigaFusionNet is a large-scale neural network architecture meticulously designed to learn a comprehensive, unified understanding of the physical world and the complex dynamics inherent to driving. This singular, powerful model ingests and processes multimodal sensor data from cameras, LiDAR, and radar to construct a holistic representation of the driving environment. This rich representation then serves as the bedrock for all subsequent critical tasks, ranging from 3D bounding boxes and 3D scene understanding to end-to-end driving token prediction. Training large-scale Physical AI foundation models like GigaFusionNet requires tightly integrated accelerated computing infrastructure optimized for multimodal AI, distributed training, and high-throughput data movement.Source: Lambda Labs — Published — Category: Models