PIMBOX is an Experimental Development project in the field of Machine Learning ("ML"). ML is an extremely active area of academic research whose exploitation to date is almost exclusively restricted to large-scale implementations such as data mining in large corporate environments and the provision of global, on-line services. Recent advances in research have shown that it is possible to reduce the scale of the computational resources required to run suitably-designed and appropriately-trained ML systems. (The importance of large and appropriate training sets is undiminished.) This holds out the possibility of bringing many of the capabilities of large-scale systems into low-cost, individualised implementations. The best-known examples are sensor signal processing for "self-driving" cars and the long-anticipated IOT. The PIMBOX project, on the other hand, challenges several of the pre-conceptions of ML-based on-line services. These require users to upload unencrypted private data to take advantage of the ML-based tools and services that are offered free-of-charge in return. "If it's free, you are the product." PIMBOX aims to provide self-trainable ML data management and curation tools for secure local home and commercial networks. These will have better usability than the best on-line services. Users may then encrypt private and personal data before using on-line backup or sharing. The project engages the close involvement of leading academics in the field of ML, with highly-experienced industrial h/w and s/w engineers, and market and business developers.