Track 08: Embeddings & Vector Databases

How text becomes a vector, what "similarity" means geometrically, how to split real documents into retrievable chunks, and how approximate nearest-neighbor search makes similarity search fast at scale. The direct prerequisite for track 09 (RAG) — most RAG failures are actually embedding or chunking failures.

Modules

1. What an Embedding Is
2. From Word2Vec to Sentence Embeddings
3. Embedding Models Compared
4. Cosine Similarity and Distance Metrics
5. The Curse of Dimensionality
6. Generating Embeddings at Scale
7. Chunking: Fixed-Size Strategies
8. Chunking: Semantic and Structural
9. Chunk Overlap and Metadata
10. The Vector Database Landscape
11. pgvector and Postgres
12. Chroma and Local Development
13. Pinecone, Weaviate and Qdrant
14. Exact vs. Approximate Search
15. HNSW Explained
16. IVF, Product Quantization and Other Indexes
17. Metadata Filtering and Hybrid Queries
18. Capstone Project