Fastagger aims to democratise AI by providing software infrastructure that allows ML and AI models to run directly on edge devices. It aims to enable multimodal LLMs, i.e. models that capture different data types, and other ML models to operate beyond the constraints of traditional cloudbased and high-performance computing systems. Fastagger primarily targets users with lower-end smartphones who are often left behind due to unreliable or limited access to mobile internet. To achieve this goal, Fastagger is currently focused on adapting pre-trained models such as OLMo, DeepSeek, LLM360, Mistral, Mosaic or Llama through compression and fine-tuning. Compression is the process of reducing the number of parameters within a model, Secteurs Renseignements artificiels Emplacement Nairobi, Kenya Etape Se connecter pour afficher les détails Marchés Se connecter pour afficher les détails Modèle client Se connecter pour afficher les détails Recettes Se connecter pour afficher les détails Prise en charge par Make-IT in Africa > Africa4Future Contact Se connecter pour afficher les détails Les médias sociaux