Built and deployed SmilhCare, a full-stack AI-driven home care platform. Developed the front-end with React and engineered a Flask/Supabase back-end API. Built a custom RAG-based chatbot from scratch using LangChain with a Qwen 2.5 7B model running on llama.cpp. Optimized inference on a CPU-only server (12 cores, 48GB RAM) to achieve 9 tokens/sec through quantization, threading, and context management. Integrated Cohere embedding models for semantic search and fast data retrieval, delivering a smooth, responsive user experience in production.