Cactus is an open-source SDK for developers to run large language, vision, and speech models directly on mobile and wearable devices cross-platform.
Cactus Compute
Cactus Compute is building the neutral standard for on-device AI. Their open-source SDK lets developers run large language, vision, and speech models directly on mobile and wearable devices cross-platform and without the lock-in of Apple or Google. By optimizing inference on specialized hardware (NPUs, DSPs) with custom kernels, Cactus enables faster, cheaper, and more private AI experiences that work offline.
We see Cactus becoming the infrastructure layer for the coming wave of edge AI or a “CUDA for smartphones”. Developers already use their SDK to power applications in healthcare, industrial devices, and consumer products where low latency and privacy are non-negotiable. We invested because the team combines world-class technical talent with sharp product execution, and they are moving quickly to establish themselves as the trusted cross-platform standard before the incumbents catch up.
Founders
Roman studied economics at University of Oxford. He is a former quant and economist with a background in product and data engineering.
Henry sidestepped a research position at Nvidia to work on Cactus. He is an incredibly strong engineer with background in AI/ML.