FORTif.ai: A Multimodal Platform for Safer Independent Living

Abstract

Global aging trends present a growing dichotomy: while older adults prefer to age in place, age-related declines in mobility and cognition can make home environments more hazardous and increase the risk of falls. This paper introduces FORTif.ai, an integrated platform that combines (1) LLM-based hazard detection to identify household risks from images and video and (2) an empathetic, LLM-powered chatbot for personalized support. The chatbot uses a retrieval-augmented generation (RAG) pipeline supported by a Weaviate vector database and a four phase workflow that incorporates hybrid search with thresholded retrieval. We present hazard detection evaluation results from preliminary investigation, across five videos, along with additional image-based testing that emphasizes key behavioral patterns and prompt-driven enhancements in detecting floor-level clutter and common tripping hazards. Collectively, these findings demonstrate a proactive approach to home safety that integrates real-time environmental sensing with conversational support tailored to older adults.