Project: Training & Fine Tuning a LLM
Executive Summary For this project, I built a fully offline, privacy-focused AI coding assistant tailored specifically for Swift development. Base AI models often lack the newest programming rules, I used a Retrieval-Augmented Generation (RAG) pipeline to feed the AI static, up-to-date Swift developer guides. This allowed me to safely bridge the gap between hardware constraints and the need for accurate, current coding assistance. Hardware Architecture Running a Retrieval-Augmented Generation (RAG) pipeline locally requires balancing compute power with strict memory constraints. The foundation of this localized AI environment relies on consumer-grade hardware optimized for parallel processing: ...