Metadata Filtering in RAG to Improve Source Relevance
Filtering metadata before vector search cuts retrieval noise at the source.
Marcus Oduya
Staff Writer
Marcus Oduya
Marcus focuses on the intersection of web data collection and machine learning pipelines, having previously built commercial scraping systems for e-commerce intelligence firms. His reporting bridges the gap between practitioner experience and editorial analysis.
8 stories
Crawl-Based Index Refresh Strategies for RAG Pipelines
Keeping RAG indexes fresh requires design choices at every pipeline stage.
Marcus Oduya
Error Handling and Fallback Strategies in Agentic Pipelines
Build recovery logic into agentic pipelines, not just better models.
Marcus Oduya
Data Normalization After Structured Web Extraction
Values inside extracted fields need standardizing before they'll work together in AI pipelines.
Marcus Oduya
Ethical and Legal Boundaries of Web Scraping for AI
Courts are still writing the rules on what AI companies can legally scrape.
Marcus Oduya
Converting Raw HTML to Clean Markdown for LLM Ingestion
Markdown cuts token bloat while preserving the structure that makes content retrievable.
Marcus Oduya
JavaScript-Rendered Content Extraction for AI Agents
AI agents silently fail on JavaScript-heavy sites until the token bill arrives.
Marcus Oduya
Schema-Driven Web Data Extraction for AI Applications
AI agents need structured, typed data from websites to act reliably without parsing text.
Marcus Oduya