The ENGRAMMER project relies on several AI technologies:
-Information is processed and represented by machine learning tools (autoencoders and transformers) over latent spaces.
-The development of the natural language interface, as well as the generation of episodes and context is supported by LLM systems.
-The protection of information, and its processing in the cloud, is supported by European projects under development in this area (AI4EOSC, SIESTA).
The platform is developed based on the initial use cases, using recurrent neural networks as autoencoders, and on systems limited in their capacity. It is therefore proposed to:
-Consolidate the secure infrastructure for storing information in the cloud.
-To develop new examples of memory consolidation and educational learning.
-Migrate from closed LLM software to an open system.
-Develop a version of the mobile application for users, allowing for scaling and general distribution through distribution systems (such as Google Play).
-Test the application in a realistic context of at least 1000 users, of different categories, in the second half of 2025, before the launch of a production version.