It’s become clear that Internet of Things (or IoT) applications work best when backed up by best-in class machine learning techniques. And between every IoT device and its machine-learning backend lies a cloud-based system. The huge volumes of data generated by these IoT devices and used by these machine learning applications have stressed current systems to their breaking point. To unlock the next stage of scaling, we are building an intelligent object store atop Derecho, our best-in class cloud computing toolkit. Our solution leverages Derecho’s RDMA-based design to introduce a zero-copy pipeline all the way from the edge to the core, all while enjoying Derecho’s rich features like automated recovery, guaranteed consistency, and explicit time-indexed versioning.