Week 11: Hybrid Search (Okapi BM25 + Dense), Reciprocal Rank Fusion (RRF), Cross-Encoder Reranking & IR Metrics (nDCG@K)

Month 3 — Enterprise Hybrid RAG, Vector Databases & Information Retrieval Goal: Build the industry-standard Two-Stage Hybrid Retrieval architecture (BM25 + Dense Vector -> RRF (k=60) -> Cross-Encoder Reranker -> Lost-in-the-Middle Context Reordering) and evaluate it with rigorous Information Retrieval metrics (Recall@K, MRR, nDCG@K).

---

Weekly Architecture & Daily Breakdown

• Day 1: Why Dense Search Fails on Exact IDs & Deriving Okapi BM25 Math (day01_why_dense_search_fails_exact_ids_and_bm25_math/)
• Day 2: Coding an Inverted Index & Okapi BM25 Search Engine from Scratch (day02_inverted_index_and_bm25_engine_from_scratch/)
• Day 3: Hybrid Search Fusion — Convex Score Combination vs Reciprocal Rank Fusion (RRF, $k=60$) (day03_hybrid_search_convex_combination_vs_rrf_math/)
• Day 4: Cross-Encoder Reranking & Solving the "Lost in the Middle" Phenomenon (day04_cross_encoder_reranking_and_lost_in_the_middle/)
• Day 5: Query Transformation — Coreference Rewriting, Multi-Query Expansion & HyDE (day05_query_transformation_rewriting_multi_query_and_hyde/)
• Day 6: Information Retrieval Metrics Math — Recall@K, Precision@K, MRR & nDCG@K (day06_retrieval_evaluation_metrics_recall_mrr_and_ndcg/)
• Day 7: Week 11 Mini-Project — Two-Stage Hybrid BM25 + Dense + RRF + Reranker Engine (day07_week11_mini_project_two_stage_hybrid_reranker/)