Hybrid Search With Dense and Sparse Vectors in RAG
Combining dense vectors with BM25 retrieval fixes the two ways RAG systems fail most often.
Priya Subramaniam
The Data Conduit — Page 4
RAG Architecture Patterns
Reranking Models as a RAG Retrieval Layer
Cross-encoders beat single-vector retrieval by rescoring candidates jointly instead of in isolation.
Marcus Oyelaran
· 12 min read
Metadata Filtering to Reduce Retrieval Noise in RAG
Filtering retrieved documents by metadata slashes hallucination rates by half or more.
Marcus Oyelaran
The Feed
Index Freshness Strategies for Live-Web RAG
Decide staleness tolerance per topic, not per pipeline.
Jailen Rafferty
Chunk Size Tradeoffs in RAG Retrieval
Marcus Oyelaran
Handling JavaScript-Rendered Pages in AI Ingestion Pipelines
Priya Subramaniam