The GEO.INFORMED project will develop generic deep learning algorithms adjusted to the EO data characteristics. Calibration and validation is an indispensable part of deep learning. We will develop generic cal/val procedures, with guidelines for ground sampling design and the re-use of existing data. The needs of its end-user community are the driving force behind the development of geo-environmental indicator. In this project, the selection of the geo-indicators and the continuous evaluation of their development will be achieved by a process of co-design. There are to this day only few examples of its use with policy makers as the end-users. The GEO.INFORMED project will thus be a test case for the use of co-design at the science-policy interface.