🔍 RAG, Vector Systems & AI Infrastructure: Sub-Curriculum Index

Welcome to the RAG, Vector Systems & AI Infrastructure curriculum. This track covers the core infrastructure of enterprise Generative AI systems: dense vector embeddings, Approximate Nearest Neighbor (ANN) index algorithms (HNSW, IVF-PQ), modular RAG pipelines, production RAG stacks, high-throughput GPU serving infrastructure (vLLM, Triton), and cross-encoder rerankers.

Every guide in this series strictly follows a two-part learning format:

  • ⚡ Quick Dive: Architecture diagrams, distance metric formulas, index trade-off tables, and Python setup snippets.
  • 📖 Extended Guide: Deep index mechanics, chunking strategies, hybrid search (BM25 + Dense + RRF), GPU memory sizing (KV cache math), and production deployment manifests.

📚 Curriculum Roadmap

# Guide Primary Topics Covered
01 Embeddings Representation & Metrics Dense semantic vector spaces, Cosine Similarity, Dot Product, Euclidean Distance, and embedding dimension sizing.
02 Vector Search: HNSW & IVF Indexing Approximate Nearest Neighbors (ANN), Hierarchical Navigable Small World (HNSW), Inverted File with Product Quantization (IVF-PQ), and Vector DBs (Qdrant, Milvus, pgvector).
03 RAG Architecture & Chunking Strategies Document parsing, sliding window vs. semantic chunking, metadata filtering, and hallucination reduction.
04 Production RAG Stack & Advanced Pipelines Complete end-to-end RAG architecture, query transformation (HyDE, Multi-Query), parent-document retrieval, and evaluation (Ragas).
05 AI Infrastructure: GPU Serving & vLLM High-throughput LLM serving with vLLM (PagedAttention), continuous batching, tensor parallelism, and Triton Inference Server.
06 Contextual Compression & Reranking Bi-Encoder vs. Cross-Encoder rerankers, Hybrid Search with Reciprocal Rank Fusion (RRF), FlashRank, and Late Chunking.
⌂ All Guides