The Euclidean Sinkhole: Why Local Dense Vector RAG Fails at Personal Knowledge SensemakingTags: RAG, Vector Database, LLM, GraphRAG, Python, System Architecture TL;DR When building Retrieval-Augmented Generation (RAG) systems over personal knowledge archives (like Obsidian vaults or deveJul 6 · 8 min readjunsukim.hashnode.dev
Spectral Graph Embeddings & Bipartite Bridging: Controlling Data Sparsity Without LLM Semantic DriftTags: Machine Learning, NLP, Graph Theory, Linear Algebra, Python, Data Science TL;DR In specialized or small-scale text domains (like personal knowledge vaults or domain-specific Korean corpora), tJul 6 · 6 min readjunsukim.hashnode.dev
