Week 10 Master Guide

Week 10: Vector Databases Internals — `IVF`, `HNSW`, Vector Quantization (`SQ8`/`Binary`) & Filtered ANN

Week 10 Daily Breakdown (7 Days)

Day 1

Week 10, Day 1: Exact Flat `kNN`, RAM/Bandwidth Math & The Curse of Dimensionality

1 code file(s) • 3 Q&A
Day 2

Week 10, Day 2: Partition-Based ANN — Inverted File Index (`IVF`), Voronoi Cells & `nlist`/`nprobe`

1 code file(s) • 3 Q&A
Day 3

Week 10, Day 3: Graph-Based ANN — Hierarchical Navigable Small World

1 code file(s) • 3 Q&A
Day 4

Week 10, Day 4: Vector Compression — Scalar (`SQ8`) & Binary (`1-Bit`) Quantization with Oversampled Rescoring

1 code file(s) • 3 Q&A
Day 5

Week 10, Day 5: Metadata & RBAC Security — Why Post-Filtering Fails (`Zero-Results Cliff`) vs Pre-Filtering

1 code file(s) • 3 Q&A
Day 6

Week 10, Day 6: Index Lifecycle — Tombstones, Versioned Upserts & Blue-Green Alias Swapping

1 code file(s) • 3 Q&A
Day 7

Week 10, Day 7 (Mini-Project): Quantized RBAC-Filtered Vector Database Engine

1 code file(s) • 3 Q&A

Week 10: Vector Databases Internals — IVF, HNSW, Vector Quantization (SQ8/Binary) & Filtered ANN

Month 3 — Enterprise Hybrid RAG, Vector Databases & Information Retrieval Goal: Understand how production vector databases (Qdrant, Milvus, pgvector, FAISS, Pinecone) search 100,000,000+ vectors in < 5 ms using IVF, HNSW, INT8/Binary Quantization, and RBAC Pre-Filtering.


Weekly Architecture & Daily Breakdown

  • Day 1: Exact Flat kNN, RAM/Bandwidth Math & The Curse of Dimensionality (day01_exact_knn_curse_of_dimensionality_and_latency_math/)
  • Day 2: Partition-Based ANN — Inverted File Index (IVF), Voronoi Cells & nlist/nprobe (day02_ivf_inverted_file_index_and_voronoi_partitions/)
  • Day 3: Graph-Based ANN — Hierarchical Navigable Small World (HNSW) (M, efConstruction, efSearch) (day03_hnsw_hierarchical_navigable_small_world_graphs/)
  • Day 4: Vector Compression — Scalar (SQ8) & Binary (1-Bit) Quantization with Oversampled Rescoring (day04_vector_quantization_pq_sq8_and_binary_quantization/)
  • Day 5: Metadata & RBAC Security — Why Post-Filtering Fails (Zero-Results Cliff) vs Pre-Filtering (day05_metadata_filtering_pre_vs_post_filtering_and_acls/)
  • Day 6: Index Lifecycle — Tombstones, Document Upserts & Blue-Green Alias Swapping (day06_index_lifecycle_upserts_tombstones_and_zero_downtime_reindexing/)
  • Day 7: Week 10 Mini-Project — Quantized RBAC-Filtered Vector Database Engine (day07_week10_mini_project_filtered_quantized_vector_store/)
Week 10: Week 10: Vector Databases Internals — `IVF`, `HNSW`, Vector Quantization (`SQ8`/`Binary`) & Filtered ANN | 24-Week AI Engineer Academy · DataArena