SiMa.ai

From a trained model to an application on the device

SiMa.ai combines its Modalix machine-learning system-on-chip with Palette software to deploy AI applications on physical devices. Developers bring models into a containerized toolchain that builds, optimizes and packages them, then use its runtime to run inference on the chip. The hardware handles sensor inputs, model execution and application logic locally, supporting uses such as object detection, robot perception and on-device conversational interfaces where power consumption and latency matter.

Equipment manufacturers and integrators can evaluate applications on a development kit, use a system-on-module or PCIe card, or integrate the chip into their own board. These are available products with a developer environment and deployment workflow. Palette Neat adds an agent workflow layer to application development. SiMa.ai’s September announcement identifies an established customer and partner network across industrial and automotive markets. Its next-generation 1,000-TOPS hardware is planned for the first half of 2028; it is separate from the current Modalix family.