Recent improvements in GNSS capabilities of various Internet-of-Things (IoT) devices, including smartphones, have allowed for better PNT performance, as well as for new potential applications in geosciences. In particular, multi-constellation and multi-frequency receivers found in recent generations of smartphones bring great potential for GNSS science exploitation. However, access to IoT data for scientific purposes is currently limited, facing different data processing challenges. The project CAMALIOT: Application of Machine Learning Technology for GNSS IoT Data Fusion addresses these issues in order to increase the usability of GNSS IoT data for scientific purposes. It encompasses the whole pipeline from raw GNSS IoT data collection and development of methods for efficient and automatic processing, to the final suitability demonstration for determination and prediction of atmospheric parameters. In this way, the project will extend the capabilities of the GNSS Science Support Centre (GSSC) of ESA, which offers GNSS data and processing services for various domains. In order to collect raw GNSS IoT data, an Android app has been developed that will be the basis of a crowdsourcing campaign. The data ingestion, processing, and analysis are designed to be highly automated, robust, and scalable. Machine learning is used for several tasks in the processing scheme, including anomaly detection as well as data fusion and prediction. The GNSS IoT data is combined with geodetic GNSS data and external models and datasets related to the atmosphere and space weather to realize global grids of tropospheric parameters (zenith wet delays, gradients, precipitable water vapor) and ionospheric vertical total electron content. This work is funded by the NAVISP Element 1 program of ESA.